FOCUS | Review Article

FOCUS | Review Article https://doi.org/10.1038/s41590-019-0399-9

https://doi.org/10.1038/s41590-019-0399-9

Adaptation and memory in immune responses

Gioacchino Natoli   1,2* and Renato Ostuni   3,4*

Adaptation is the ability of cells, tissues and organisms to rapidly and reversibly modify their properties to maximize fitness in
a changing environment. The activity of immune-system components unfolds in the remarkably heterogeneous milieus to which
they are exposed in different tissues, during homeostasis or during various acute or chronic pathological states. Therefore,
adaptation is essential for immune cells to tune their responses to a large variety of contexts and conditions. The adaptation of
immune cells reflects the integration of multiple inputs acting simultaneously or in a temporal sequence, which eventually leads
to transcriptional reprogramming and to various functional consequences, some of which extend beyond the duration of the
stimulus. A range of adaptive responses have been observed in both adaptive immune cells and innate immune cells; these are
referred to with terms such as ‘plasticity’, ‘priming’, ‘training’, ‘exhaustion’ and ‘tolerance’, among others, all of which can be
useful for defining a certain immunological process or outcome but whose underlying molecular frameworks are often incom-
pletely understood. Here we review and analyze mechanisms of adaptation and memory in immunity with the aim of providing
basic concepts that rationalize the properties and molecular bases of these essential processes.

I
mmunological memory is defined as the ability of immune cells
to specifically ‘remember’ the first encounter with a given anti-
gen and to mount a secondary response that is faster and greater
in magnitude than the primary response. The specificity of these
memory responses is very high and is based on irreversible changes
in the DNA sequence during and after rearrangement of the antigen
receptors on B lymphocytes and T lymphocytes. A critical contribu-
tion to the stability of memory responses is provided by changes
in DNA methylation (notably, passive demethylation) that can be
transmitted to daughter cells during the clonal expansion of anti-
gen-stimulated naive lymphocytes and thus enable the fast and effi-
cient re-activation of genes encoding critical effector molecules in
memory cells1,2.
Immunological adaptation is instead the process whereby leuko-
cytes exposed to environmental stimuli modify their properties in
a way that influences future responses to the same or other stimuli.
The adapted state is reversible and dynamic, which means that it
can be resolved when environmental conditions return to the initial
state and/or it can be modified in response to subsequent micro-
environmental changes. As discussed below, the reversibility of
adaptive states is linked to specific mechanisms that drive molecu-
lar alterations that can be rapidly ‘erased’ even in non-dividing cells,
such as those that occur at the chromatin, transcriptional, post-tran-
scriptional or metabolic level. Although they are reversible, adaptive
phenotypes and the associated molecular changes may in some cases
persist for variable time windows (up to weeks or even months)3,4.
Hence, the perpetuation of adaptation in some contexts, particularly
in innate immunity, has been considered a form of immunological
‘memory’ (‘trained immunity’ or ‘short term memory’)3,4.
Conceptually, the degree of reversibility of an acquired prop-
erty of immune cells should define the border between sustained
adaptation and true memory (Fig.  1). The other critical feature
that distinguishes memory from adaptation is the specificity of the
response. Memory is specific by definition. Conversely, an adaptive
process can be pervasive or global in nature. Indeed, while memory
lymphocytes are selectively activated by cognate antigens, adapted
myeloid cells can display a general state of hyper- or hypo-respon-
siveness to a broad set of unrelated stimuli5.
Because memory implies specificity, we propose to restrict the
use of this term to antigen-specific immunity and to consider the
long-lasting effects of cell stimulation in the immune system as
forms of adaptation characterized by slow reversibility (Fig.  1).
Notably, lymphocytes also show adaptive responses commonly
referred to as ‘plasticity’: the ability of a given lymphocyte to func-
tionally adapt its effector responses to a changing environment.
Along the same lines, T cell exhaustion can be considered a form
of adaptation characterized by remarkable stability. Therefore,
while bona fide immunological memory is a property of lympho-
cytes, adaptation is a property of essentially all immune-cell types
(as well as non–immune-cell types), with various degrees of stabil-
ity but without the specificity of true memory. The peculiar case of
memory in natural killer (NK) cells has been discussed in excellent
reviews6,7. In this specific case, the enhanced responsiveness that
follows the initial activation and clonal expansion phase can persist
for months8 and resembles memory in the adaptive system (possibly
also from a mechanistic point of view, as discussed below), with the
obvious difference of the innate-type, limited specificity of NK cell
responses. In the next paragraphs, we will focus on the adaptation
of immune cells, by first defining the properties of the responses to
various acute and chronic stimuli and then by discussing mecha-
nisms that underlie those processes.
Properties of eliciting stimuli affect adaptive phenotypes
Adaptations in the immune system are extremely diverse in terms of
biological output, and they affect future responses and mechanisms
that control the establishment and maintenance of the adapted
state (Fig.  2). This diversity is a direct reflection of the multitude
of stimuli that can modify the activity of immune cells as well as
of additional quantitative properties of environmental signals, such
as their intensity or persistence. It is well known that while distinct
stimuli typically elicit specific adaptations, a given signal may trig-
ger very different types of adaptations if it is present at a low or high
dose or if it is sensed for a short or long period of time. For example,
repeated or persistent exposure of macrophages to high doses of
inflammatory agents, such as the microbial component lipopoly-
saccharide (LPS), often elicits a state of hypo-responsiveness that
prevents further re-induction of genes encoding inflammatory mol-
ecules (called ‘inflammatory genes’ here)5. On the other hand, very
low doses of LPS or other inflammatory stimuli are generally linked
to a potentiation of the response to future challenges3.

1
Humanitas University, Pieve Emanuele, Milan, Italy. 2IRCCS Humanitas, Rozzano, Milan, Italy. 3San Raffaele-Telethon Institute for Gene Therapy (SR-Tiget),
IRCCS San Raffaele Scientific Institute, Milan, Italy. 4Vita-Salute San Raffaele University, Milan, Italy. *e-mail: gioacchino.natoli@hunimed.eu;
ostuni.renato@hsr.it

Nature Immunology | www.nature.com/natureimmunology

Review Article | FOCUS
https://doi.org/10.1038/s41590-019-0399-9
Adaptation Memory
Reversibility Yes No
Short-term or
Persistence
long-term
Long-term
Specificity No Yes
• Receptor signaling/recycling • DNA sequence alterations
• Metabolic reprogramming • Epigenetic modifications
• Chromatin/histone (DNA methylation)
Mechanisms modification • Self-sustaining feedback
• TF occupancy/distribution loops
• Induction of long-lived
mediators
Innate, adaptive immune cells
Target cells
Non-immune cells
Adaptive immune cells
Fig. 1 | Discriminative properties of adaptation and memory in the
immune system. Adaptation to and memory of environmental stimuli
should be defined mainly by the degree of reversibility or persistence of
the elicited state and by the specificity of the response after secondary
stimulation. In this framework, adaptation is controlled by molecular
mechanisms that involve reversible biochemical, metabolic or chromatin
alterations that are diluted across cell division, while memory relies on
stable genetic or epigenetic changes, as well as on self-sustaining loops.
Because almost all types of cells are equipped with molecular machinery
able to trigger reversible changes, adaptation can be considered a
universal cellular property. Enzymes (such as the RAG recombinases or
the activation-induced cytidine deaminase AID) that can elicit irreversible
changes are instead active only in cells of the adaptive immune system,
which underlies memory in this compartment.
The divergent outputs of possible adaptations to a common
stimulus probably reflect the associated functional requirements:
preventing immune system–mediated tissue damage and excessive
inflammatory responses on the one hand and enhancing immune
responses to concomitant or future threats on the other. In addi-
tion to modulating the outputs of the adapted states, the intensity
of stimulation can underlie the extent of the persistence and revers-
ibility of the states. Transient and/or low-dose stimulation typically
elicits reversible adaptations that may be resolved in short time
frames (hours to days) or long time frames (weeks to months); this
is exemplified by the short-term priming effect of inflammatory
signals on macrophages (described below) or by the long-term, but
flexible, adaptations of CD4+ helper T cells after transient exposure
to polarizing cytokines9. On the other hand, persistent and/or high-
dose stimulation often elicits persistent adaptations that, to some
extent, may be not easily, rapidly or completely reverted. For exam-
ple, chronic antigen stimulation of CD8+ T cells leads to a state of
exhaustion that is associated with stable transcriptional reprogram-
ming and acquisition of specific functional features, which can be
reverted only partially after blockade of inhibitory immunological
checkpoints10,11.
Adaptation to tissue contexts
A common output of adaptation is the acquisition of context-
dependent activities, as exemplified by the tissue-specific functions
of immune cells and of macrophages in particular12,13 (Fig. 2). The
diverse activities of tissue macrophages are enabled by specific tran-
scriptional programs that reflect differential selection and usage of
the available cis-regulatory repertoire (specifically, the complement
of accessible genomic regulatory elements) in response to tissue-
specific cues14,15. Mechanistically, the first layer of genomic regu-
lation is provided by myeloid lineage–determining transcription
factors such as PU.1, IRF8, C/EBPα, C/EBPβ and others, which act
in a combinatorial fashion during differentiation to establish and
maintain the basal and inducible cis-regulatory landscape of mac-
NATure Immunology
rophages16–19. With some notable exceptions, represented by latent
or de novo enhancers20,21, the set of enhancers made accessible by
lineage-determining transcription factors serves as a platform onto
which stimulus-responsive transcription factors land after activa-
tion, which thus links cell identity to activation14,15. Tissue-resident
macrophages are exposed to cues that are linked to the specific prop-
erties of the organ, such as heme in the spleen, retinoic acid in the
peritoneum or the microbiota in the intestine13. Such signals in turn
stimulate the activity of tissue-restricted transcription factors, such
as GATA-6 in the peritoneum22–24, SPI-C in the spleen25, LXRα and
ID3 in the liver26,27 and many others, which control the selection of
tissue-specific enhancer landscapes and transcriptional programs in
the tissue-resident macrophages28,29. Despite being conditioned by
local cues, the regulatory repertoire of resident macrophages, and
eventually the tissue-specific transcriptional output of these cells, is
to a large extent plastic and reversible, as indicated by experiments
in which macrophages are transferred from one tissue to another28.
The distinct genomic organizations of resident macrophages
represent paradigmatic adaptations that allow context-dependent
responses to stimulation. For example, microglia have high expres-
sion of transcription factors such as MEF2C30, MAFB31 and NR4A132,
which act to limit the induction of inflammatory genes in response
to stimulation and thus prevent neuronal damage. Expression of
the transcription factor SPI-C in some tissue macrophages, such as
spleen macrophages, Küpffer cells and intestinal macrophages, sup-
presses induction of the inflammatory cytokines interleukin 6 (IL-6)
and IL-1α (but not that of TNF); here, SPI-C acts by interfering with
interaction of the transcription factor NF-κB subunit p65 with the
transcription factor IRF5 in response to LPS33. Conversely, GATA-6
inhibits the expression of genes encoding molecules linked to type 2
immunity, such as Mrc1, Cd163, Arg1 and Chi3l323, which indicates
that this transcription factor might boost inflammatory responses
in peritoneal macrophages, whose activation reflects critical emer-
gencies that require rapid and effective responses. Emerging evi-
dence indicates that the ability to adapt to local signals is a general
feature of immune cells, although it is unclear whether the underly-
ing molecular mechanisms follow the same rules as those described
above for macrophages. For example, neutrophils infiltrate tissues at
steady state34 and display phenotypic and functional heterogeneity35.
Phagocytosis of neutrophils in the bone marrow regulates circadian
oscillations of the number of mobilized hematopoietic progenitor
cells36, while their phagocytosis in the intestine suppresses sys-
temic levels of the cytokine G-CSF and controls overall bone mar-
row activity34. Additional tissue-specific functions of neutrophils
include the support of B cell maturation in the spleen37 or the regu-
lation of diurnal transcriptome changes in the lungs34. Along the
same lines, conventional T lymphocytes, innate lymphoid cells and
NK cells show tissue-specific profiles and functions38.
Even if located within the same tissue, immune cells may still
function differently according to their ontogeny, as they often rep-
resent the progeny of distinct hematopoietic precursor cells and
include cell populations that are maintained independently from
adult hematopoietic stem cells (HSCs)39. This is now well estab-
lished for tissue-resident macrophages40, as well as for mast cells41,42,
and for lymphocyte populations such as innate-like B-1 cells43, γδ
T cells44 or CD8+ T cells45. Emerging evidence indicates that ontog-
eny affects how immune cells interact with the environment, which
in turn implies that different regulatory networks operate in cells
of different origin, even if the cells are exposed to the same tissue
microenvironment. For example, adult monocyte-derived macro-
phages that engraft the brain post-natally acquire transcriptional
and chromatin features that only partially resemble those of yolk
sac–derived microglia and display different capacities to respond
to inflammatory insults46,47. Distinct molecular organization and
functional properties have been reported for monocyte-derived
macrophages and embryonic resident macrophages in pancreatic
Nature Immunology | www.nature.com/natureimmunology
 FOCUS | Review
| Article
NATure Immunology FOCUShttps://doi.org/10.1038/s41590-019-0399-9
Review Article
NF-κB NF-κB STAT1
Ac Ac
Hyper-responsive
a Termination of stimulus
adaptation

Induction of signaling components

Induction of transcription factors (e.g., STAT1)
Metabolic reprogramming
1st stimulus Higher chromatin accessibility 2nd stimulus
(LPS) (e.g., LPS)
Receptor desensitization NF-κB
Induction of repressive factors
(e.g., miR-221/miR-222)
NF-κB Metabolic reprogramming Me3
Ac
Lower chromatin accessibility
Hypo-responsive
adaptation
Persistence of stimulus

TF A NF-κB TF A
b Ac

Poised Active
Ac
Active Active

Inactive Inactive

Inactive (e.g., tissue signal A)

Context-dependent
Inactive 2nd stimulus adaptation
1st stimulus
(e.g., LPS)
Inactive
(e.g., tissue signal B) NF-κB
Me3

Repressed
Repressed
Inactive
Inactive

Active
Active
TF B
TF B
Selection of tissue-specific enhancers
Tissue-specific response to stimulation
Tissue-specific basal gene expression

Fig. 2 | Different outputs of immunological adaptations dictated by the properties of the eliciting stimulus. a, A hypothetical locus of inflammatory
genes (blue) in macrophages showing hyper-responsive adaptations (top) or hypo-responsive adaptations (bottom) elicited by a transient (low-dose)
exposure or persistent (high-dose) exposure, respectively, to an inflammatory agent. Transient inflammatory stimulation of macrophages, such as with
LPS, triggers complex changes at the signaling, metabolic or transcriptional level that often result in increased histone acetylation (Ac) and greater
chromatin accessibility at regulatory elements of adapted genes, which in turn favors the binding of transcription factors and gene expression after
secondary exposure to another inflammatory stimulus (‘priming’ or ‘training’). On the other hand, persistent inflammatory stimulation of macrophages
often elicits negative feedback loops that lead to a reduction in signaling strength, induction of anti-inflammatory factors, such as miR-221 and miR-222,
and metabolic rewiring. Such alterations are associated with reduced chromatin accessibility and, in some cases, deposition of repressive histone marks
such as H3K27me3 (Me3) at regulatory elements of inflammatory genes, which prevents binding of transcription factors and efficient locus re-induction
(‘LPS tolerance’). b, Hypothetical context-dependent adaptation at a locus of inflammatory genes (blue) in tissue-resident macrophages. After exposure
to local cues, tissue-specific transcription factors control the selection and activation of the cis-regulatory landscape (orange or purple loci) in resident
macrophages. Transcription factor A (TF A; orange), activated by tissue-associated signal A (tissue signal A), also binds to regulatory elements of an
inflammatory gene and primes chromatin at this locus, while transcription factor B (TF B; purple), activated by tissue-associated signal B (tissue signal
B), promotes deposition of repressive histone marks at this locus. The differential chromatin organization at this same locus, imparted by the distinct
transcription factors in tissue A versus tissue B, underlies the different transcriptional outputs in resident macrophages that lead to context-dependent
adaptations.

cancer48 and glioblastoma49. Similarly, CD8+ T  cells that originate
during fetal development or adult development display distinct
transcriptional and chromatin landscapes, which underlie differ-
ential effector capacities and dynamics45. Such observations clearly
indicate that the ontogeny of immune cells is associated with persis-
tent molecular properties that are only partially reversible through-
out the cellular life cycle. While the underlying mechanisms remain
largely unknown, a sound hypothesis is that the source of progeni-
tor cells (such as primitive HSCs versus definitive HSCs39) may
impart stable epigenetic features — specifically, changes in DNA-
methylation profiles at regulatory elements of key genes — which
results in distinct adaptations to identical tissue environments.
Hyper-responsive adaptation
Macrophages acutely exposed to LPS upregulate hundreds of
transcripts in a highly dynamic and coordinated fashion50,51.
Among those, a limited set of transcripts originally referred to as
‘non-tolerized’ transcripts are induced at higher levels and with
faster kinetics after re-challenge with LPS 24 hours after the pri-
mary stimulation52. That set of genes shows enrichment for genes

Nature Immunology | www.nature.com/natureimmunology

Review Article | FOCUS
https://doi.org/10.1038/s41590-019-0399-9
Signal response Signal integration
State A epithelial stem cells that, after exposure to the innate immunore-
Stimulus
A
B as for lung basal cells, in which adaptations to IL-4 or IL-13 drive
Stimulus A
B
B Hypo-responsive adaptation
C
C ?
C
Naive D ther challenges (Fig. 2). This reflects the existence of auto-inhibitory
macrophage
D D
E immune system–mediated pathology. For example, macrophages
E high dose of LPS acquire a hypo-responsive state, known as ‘LPS
E
Principles of inducible gene expression Principles of context-dependent
gene expression
Fig. 3 | Signal-response and signal-integration models of macrophage
activation. In the current model of macrophage activation in response to
environmental signals (left), exposure to diverse stimuli elicits specific
molecular changes and functional states. Analyses of the transcriptional
and epigenomic effects of individual stimuli have provided fundamental
information on the principles governing inducible gene expression50,51,
as well as on the diversity of activation states of macrophages95,96. The
model of signal integration at right shows the ability of macrophages to
integrate multiple and even contrasting environmental signals into specific
genomic and functional programs. Mechanistic understanding of signal
integration are expected to yield new paradigms of context-dependent gene
expression.
encoding anti-microbial effector molecules, which suggests that this
type of adaptation may boost the protective actions of macrophages
directed against infection while preventing the sustained expression
of typical inflammatory genes, whose products might cause severe
damage5. Potentiated re-induction of specific sets of genes has been
reported in macrophages or fibroblasts after repeated exposure to
diverse cytokines such as interferon-γ (IFN-γ)20, IFN-β53 or IL-420,54.
The state of heightened responsiveness acquired by macrophages in
inflammatory conditions often extends to secondary exposure to
related, but not identical, signals (Fig. 2). For example, pre-exposure
of macrophages to IFN-γ boosts the expression of many inflamma-
tory genes after subsequent stimulation with LPS55; innate immune
cells stimulated with β-glucans, key components of the fungal cell
wall, mount more-efficient inflammatory responses to a variety of
bacterial pathogens in vitro and in vivo3.
Mechanistically, increased expression of inflammatory genes
after re-stimulation of macrophages relies on various processes,
including increased expression of signaling-pathway com-
ponents, metabolic reprogramming, such as the mTOR- and
HIF1-α-dependent glycolytic switch induced by β-glucans56,
and the deposition at cis-regulatory elements of histone modi-
fications, whose persistence after removal of the initial stimu-
lus might facilitate or potentiate secondary activation20,52,57. In
principle, therapeutic strategies that promote adapted states of
hyper-responsiveness in innate immune cells could be exploited
to boost vaccination efficacy as well as to promote anti-tumor
immunity. On the other hand, this type of adaptation may also
exacerbate immune reactions in chronic inflammatory diseases
or in autoimmunity. Repeated systemic administration of low
doses of LPS boosts the expression of inflammatory genes in
brain macrophages, which leads to exacerbated neuropathology58.
NATure Immunology
Interestingly, immunological stimuli can elicit hyper-responsive
adapted states in non-immune cells as well, as reported for skin
ceptor TLR7 or activators of the NLRP3 inflammasome, gained
regenerative capacities after subsequent tissue damage59, as well
barrier-tissue dysfunction during allergy60.
Persistent or high-dose stimulation of immune cells often elicits
states of adaptation characterized by reduced responsiveness to fur-
loops aimed at preventing excessive inflammatory responses and
exposed to sustained stimulation with LPS or stimulation with a
tolerance’, in which a large number of inflammatory genes are not
efficiently re-induced after secondary stimulation with LPS or other
inflammatory cytokines5. The blunted immune reactions by cells
that acquired a hypo-responsive state are associated with pathologi-
cal conditions such as chronic infection, in which prolonged anti-
gen stimulation of T lymphocytes causes a gradual decrease in T cell
effector functions and unresponsiveness to stimulation, a state often
referred to as ‘exhaustion’61. Analogously, the hypo-responsiveness
of innate immune cells in patients with sepsis underlies immune-
system paralysis, a state of immunosuppression that persists after
clearance of the pathogen and is associated with a higher risk of
secondary infection, organ failure and mortality in the first weeks
after resolution of sepsis62.
A variety of mechanisms have been shown to contribute to the
LPS-tolerant state of macrophages, including receptor desensiti-
zation, secondary activation of negative signaling modules and
induction of anti-inflammatory microRNAs5. In this context, the
organization and dynamics of chromatin at regulatory elements of
inflammatory genes seem to be relevant to LPS tolerance. The lack
of re-induction of tolerized genes correlates with defective depo-
sition of histone marks associated with active transcription, chro-
matin accessibility and nucleosome remodeling at promoters and
enhancers of tolerized genes52,57. On the one hand, such defective
deposition of histone marks at tolerized genes might reflect the
lower signaling strength observed after re-stimulation of toler-
ized macrophages. On the other hand, chromatin organization at
the promoters of tolerized genes differs in several ways from that
of non-tolerized genes even in unstimulated macrophages, which
suggests that the propensity for or refractoriness to tolerization may
be an intrinsic property of LPS-responsive genomic regulatory ele-
ments52. In this context, the capacity of some stimuli such as inter-
ferons or β-glucans to revert tolerance63,64 might be related to their
ability to induce chromatin changes that make some inflammatory
genes capable of responding efficiently to the diminished signals
observed in tolerized cells and thus in fact compensate for the lower
signaling strength65. Along the same lines, non-tolerized genes
— specifically, genes that are efficiently re-induced by secondary
stimulation of tolerized macrophages — may be characterized by
a chromatin state that enables efficient activation in response to
weaker upstream signals.
Mechanisms for the persistence of adaptive states
Most adaptations in the immune system are plastic and are thus rap-
idly reversible. However, in specific cases, these states can persist
and display memory-like features. In principle, persistent adapta-
tion to a transient stimulation may be acquired and maintained via
at least three non–mutually exclusive mechanisms.
The first mechanism is the activation of a self-sustaining positive
feedback mechanism able to independently propagate the initial
response, even after removal of the eliciting stimulus. Mathematical
Nature Immunology | www.nature.com/natureimmunology

NATure Immunology
models have shown that memory based on transcriptional circuits
can typically develop when the system is ‘bi-stable’; that is, when, in
response to stimulation, the system can move from one stable state
to an alternative, self-maintained, stable state66. This memory is
thus restricted to systems that can exist in only two main states (for
example, on and off, or active and repressed), such as the lac operon,
which can be present in an active state or in a repressed state, and
the phage lambda, which exists in a lysogenic phase (in which the
prophage genome is repressed) and a lytic phase (in which genes
encoding molecules required for replication are active)67. The exis-
tence of such bi-stable switches in the context of immune-cell adap-
tation has never been reported, to our knowledge.
A second mechanism, the induction of long-lived mediators,
includes the expression of microRNAs, whose half-life can be
extremely long and result in long-term cellular effects. An interest-
ing example is provided by the microRNAs miR-221 and mi-R222,
which are involved in LPS tolerance68. These microRNAs negatively
control the expression of Brg1, a catalytic subunit of the BAF (SWI/
SNF) chromatin-remodeling complex. Reduced expression of Brg1
prevents efficient re-induction of those inflammatory genes that are
highly dependent on SWI/SNF-mediated nucleosome remodeling,
such as the gene encoding IFN-β69. In keeping with this observation,
genes tolerized by the miR221–miR222–Brg1 axis include many
genes encoding interferon-stimulated transcripts, as a consequence
of defective autocrine and paracrine IFN-β signaling and secondary
activation of the STAT and IRF families of transcription factors68.
An extension of this mechanism is the evolution of the response
over time due to the reduced concentration of some mediators and
the increased concentration of others, as exemplified by coordinated
and dynamic changes in transcription factors of the REL–NF-kB,
AP-1, STAT and IRF families over many hours after initial stimula-
tion. For example, interferons induce the expression of STAT1 and
STAT2, which can potentiate responses at re-stimulation. An inter-
esting category of potential mediators of persistent adaptation is
rare proteins with exceptionally long lifespans, although their pos-
sible effect on immune responses is largely unknown70.
A third mechanism is a stable and self-propagating molecular
change, typically a change in DNA methylation that can be trans-
mitted across subsequent DNA syntheses because of coupling of
the DNA-methylation machinery with the replication fork. In the
context of typical memory responses in lymphocytes, changes in
DNA methylation provide a critical contribution. Specifically, the
proliferation of antigen-stimulated naive lymphocytes provides an
opportunity for the passive loss of repressive methyl groups at CpG
dinucleotides, if the modification on the parental DNA strand is not
copied onto the daughter strand. This is the mechanism by which
genes encoding critical effector cytokines undergo enhanced acti-
vation in memory cells. Loci such as those encoding IL-4, IL-13 or
IFN-γ are methylated in naive cells and undergo passive demeth-
ylation after activation, which thus enables stronger and faster
responses at re-stimulation1,2. Changes in DNA methylation may
also contribute to the persistence of adaptation, as shown in T cell
exhaustion, in which de  novo deposition of DNA methylation is
required for the maintenance of T  cell hypo-responsiveness71. We
are tempted to speculate that dilution of DNA methylation at criti-
cal regulatory elements after the proliferation of progenitor cells
might contribute to persistent changes in differentiated cells during
emergency myelopoiesis72 and thus might account for the mainte-
nance of the observed effects. Along the same lines, passive dilu-
tion of DNA methylation might contribute to memory in the NK
cell system, as NK cell memory responses are preceded by a phase
of clonal expansion8. Adaptive responses in proliferating cells (NK
cells and hematopoietic progenitor cells) and non-proliferating cells
might thus be critically different, because DNA demethylation dur-
ing replication provides the opportunity for the efficient propaga-
tion of a bona fide, long-lasting epigenetic change.
Nature Immunology | www.nature.com/natureimmunology
FOCUS | Review
| Article
FOCUShttps://doi.org/10.1038/s41590-019-0399-9
Review Article
The role of histone modifications
The current evidence of memory-like, long-lasting adaptive phe-
notypes in the innate immune system has been linked to the per-
sistence of stimulus-induced histone modifications associated with
gene activation3. Given the current knowledge on the turnover of
nucleosomes and their modifications, the amino termini (‘tails’)
of nucleosomal histones are subjected to a large number of modi-
fications, usually characterized by a very fast turnover (minutes
to hours), determined by the antagonistic activity of ‘writer’ and
‘eraser’ enzyme pairs, such as histone acetyltransferases and histone
deacetylases73. Some of these modifications, notably acetylation, act
to physically relax the chromatin fiber and thus make the underly-
ing DNA more accessible and prone to transcription, while others,
such as histone methylation, act as platforms that might favor or
stabilize the recruitment of various enzymatic complexes that con-
trol chromatin organization, transcription and DNA replication. In
addition to the rapid turnover of histone modifications determined
by erasers (such as deacetylases and demethylases), nucleosomes
themselves are rapidly turned over by dedicated ‘machines’ at active
genomic regions74–76.
The highly dynamic nature of nucleosomes, which represents the
basis for restoration of the basal state in a continuously changing
micro-environment, is an obvious barrier to the long-term main-
tenance of a stimulus-induced chromatin state, such as a gain in
acetylation at an activated promoter or enhancer. The gene-specific
mechanisms that overcome such high turnover rates have not yet
been identified, although some genomic regions might display an
intrinsically different ability to undergo erasure events; a slower era-
sure rate might thus result in the short-term persistence of newly
deposited histone modifications20. Among known histone modifi-
cations, the only one shown to be associated with a maintenance
mechanism similar to that of DNA methylation is the trimethyl-
ation of histone H3 at Lys27 (H3K27me3), a repressive modifica-
tion77. Conversely, there is currently no mechanistic support for the
notion that in mammalian cells, modifications such as acetylation
and activation-associated methylations can be propagated across
DNA synthesis or can be self-sustained for a long period of time.
In the absence of such a mechanism, or until it is discovered, the
long-term persistence of a stimulus-induced activation-associated
histone modification, such as trimethylation of histone H3 at Lys4
(H3K4me3) in response to stimulation with β-glucans78, might be
accounted for by two mechanisms: the maintenance of a transcrip-
tion factor–dependent regulatory circuit that self-sustains after
stimulus termination and continues to promote the deposition of
active histone modifications at target genes, or incomplete removal
of the stimulus.
Sustained adaptation of the innate immune system
Probably the best-known case of sustained adaptation in the
immune system is trained immunity: the short- to medium-term
(weeks to months) persistence of a state of enhanced responsiveness
to microbial challenges of innate immune cells previously exposed
to an initial stimulation4. The existence of such a hyper-responsive
state of the innate immune system is supported by a plethora of epi-
demiological and experimental evidence that has been discussed in
detail elsewhere3. One mechanistic issue that has not received due
attention in this context is the role of the persistence of the trigger
of a trained-immunity phenotype.
The two most extensively studied triggers of trained immu-
nity are β-glucans and mycobacteria, notably the bacillus
Calmette-Guérin (BCG) used for vaccination against tuberculosis.
Chemically, β-glucans are polymers that share with cellulose their
basic component, β-glucose, but differ from cellulose in the nature
of the chemical bonds that link the β-glucose units79. β-glucans are
structural components of fungi (such as Candida albicans), yeast,
edible mushrooms and some plants (mainly seaweed) and show
Review Article | FOCUS
https://doi.org/10.1038/s41590-019-0399-9 NATure Immunology

a Synergy b Antagonism

IFN-γ IFN-γ

Ac Ac Me3
STAT1 STAT1

LPS IL-4

STAT6
NF-κB

IFN-γ
IFN-γ LPS IL-4

Synergized locus Antagonized locus A Antagonized locus B

Fig. 4 | Chromatin control of signal integration in macrophages exposed to synergistic or antagonistic stimuli. a, Stimulation of macrophages with
IFN-γ or LPS alone (top and middle) triggers partial chromatin accessibility and limited transcriptional induction of a hypothetical locus of inflammatory
genes (blue). After combined stimulation (bottom), transcription factors induced by the two stimuli act together at these regulatory elements to increase
accessibility and promote the deposition of histone acetylation and of other features of permissive chromatin, which leads to potent gene expression of the
synergized locus. b, In addition to eliciting chromatin accessibility and transcriptional induction at target genes, IFN-γ and IL-4 (signals with antagonistic
biological outputs) actively promote de-activation of enhancers and promoters that control reciprocal gene targets. While the underlying molecular
mechanisms are not entirely clear (as discussed in the text), the net effect of this antagonism is that combined macrophage stimulation with IFN-γ plus
IL-4 leads to broad and selective impairment of both IFN-γ target genes and IL-4 target genes. Additional gene-specific activities can be observed in these
experimental settings100–102, including genes whose expression is not antagonized or is even synergized by the antagonistic stimulus, but are not depicted
here.

extreme chemical heterogeneity, from linear molecules to highly
branched molecules with a different number of subunits and a vari-
able percentage of branching80. The most relevant β-glucans for the
induction of trained immunity are those with (1→3)-β-d-glycosidic
bonds in the linear polymer and side chains bound by (1→6)-β-d-
glycosidic linkages, which are all recognized by the pattern-recogni-
tion receptor dectin-181,82. The immune system–stimulating effects
of β-glucans vary greatly on the basis of the number of units and the
branching patterns, but 11 units are in general sufficient to trigger
macrophage activation83.
Although they are chemically heterogeneous, such bioactive,
immune system–stimulating polysaccharides share the prop-
erty of not being degraded in mammals, which lack β-glucanase
enzymes. In mammals, the only pathway available for their degra-
dation is the uptake by macrophages, in which they are subjected
to a slow oxidative process84. Such slow oxidation explains the
retention of β-glucans within macrophages for extended periods
of time after internalization85. The lack of highly active, dedicated
degradation pathways explains the persistence of intravenously
injected (1→3)-β-glucans in the liver and spleen of mice for more
than 1 month (refs. 86,87), although it is unclear whether (or for
how long) these persistent β-glucans retain their biological activ-
ity. The persistence of β-glucans in cells and tissues raises the pos-
sibility that one critical component of trained immunity, at least
in some contexts and/or in response to some inducers, is the per-
sistence of the trigger, which would create a long-lasting primed
state due to low-level, continuous triggering of immune cells. This
condition of sustained exposure to an innate-immune-response
trigger would be evocative of the persistent hyper-activation state
induced by super-low concentrations of LPS, such as those that are
frequently detected in asymptomatic people and in patients with
chronic diseases88–91.
Similar considerations could apply to mycobacteria. BCG vac-
cination induces T  cell–independent protection against unrelated
pathogens, including fungi and parasites3. Because BCG is a live
attenuated vaccine, its persistence in the host may represent an
important factor in the generation of sustained training of the innate
immune system92,93. Moreover, many of the lipids of mycobacteria,
complex structures critical for the various phases of their life cycle
and in many cases endowed with biological activity94, are extremely
difficult to digest; therefore, their persistence after the clearance of
live bacteria might generate long-lasting priming effects.
Signal integration as driver of adaptive responses
The current knowledge of how cell activity is influenced by envi-
ronmental signals is built substantially on the effects of individual
stimuli on cell populations, as exemplified by refined models of
macrophage activation95. However, it is clear that the cellular and
molecular effects of individual stimuli can be heavily influenced by
co-existing environmental cues, such as those that are acutely co-
delivered to cells or that are part of the local milieu in which cells
are embedded96. We refer to this feature as ‘signal integration’, mean-
ing the ability of cells to integrate multiple and complex inputs into
unique and specific outputs that underlie the multifaceted functions
of immune cells in homeostasis and disease (Fig. 3). As discussed
below, the possible outputs of signal integration range from neu-
trality, when two or more stimuli do not influence each other, to
reciprocal modulation, synergism or antagonism. Signal integration
in immune cells occurs at multiple levels, including signal-trans-
duction pathways and especially genomic cis-regulatory elements,

Nature Immunology | www.nature.com/natureimmunology


NATure Immunology
where multiple transcription factors regulated by lineage- and stim-
ulus-specific signals land in close proximity (Fig. 4).
Combined exposure to suboptimal doses of functionally coher-
ent stimuli (for example, inflammatory stimuli such as TNF and
IFN-γ) often leads to the potentiation of cellular responses. Signal
integration in this context is key to tailoring the potency of immune
reactions to the context in which cells are activated. For example,
combined sensing of microbial and damage-associated signals, such
as ATP and many others, by innate immune cells leads to increased
production of the cytokine precursor pro-IL-1β, inflammasome
activation and, eventually, increased secretion of mature IL-1β97.
Along the same lines, dendritic cells exposed to LPS together with
oxidized phospholipids from the host, which are released by dying
cells, enter a state of hyper-activation with more secretion of IL-1β
and powerful stimulation of adaptive immunity98. A paradigmatic
example of synergism in the innate immune system is the ‘priming’
effect of IFN-γ: its ability to increase the production of inflammatory
cytokines by macrophages exposed to other stimuli, such as Toll-like
receptor agonists55. Mechanistically, stimulation of human macro-
phages with IFN-γ increases the accessibility of regulatory elements
of inflammatory genes, via STAT1 and IRF1, and thus allows greater
transcription of these genes in response to LPS55,99. Other inflam-
matory cytokines, such as TNF and IFN-α, have also been shown
to increase chromatin accessibility, transcription-factor occupancy
and deposition of H3K4me3 at promoters of a specific set of inflam-
matory genes, which leads to amplified transcriptional responses
after stimulation with a suboptimal concentration of LPS65.
Signal integration at the chromatin level also seems to govern
antagonistic responses in cells exposed to stimuli with oppos-
ing biological functions, such as pro-inflammatory cytokines and
anti-inflammatory cytokines. Genomic analyses of macrophages
co-exposed to antagonistic signals such as IFN-γ and IL-4 or LPS
and IL-4 have revealed a selective transcriptional antagonism that
leads to the defective induction of dozens of genes from the pro-
grams elicited by each stimulus alone100–102. Reciprocal repression
occurs in the absence of detectable interference with proximal sig-
naling but is linked to altered chromatin remodeling at regulatory
elements of antagonized genes100–103. Treatment with IL-4 elicits loss
of histone acetylation and of chromatin accessibility at promoters
and enhancers of inflammatory genes, which are induced at lower
levels after concomitant or subsequent stimulation with IFN-γ or
LPS100,102. While this effect is almost entirely dependent on STAT6,
only a minority of the antagonized enhancers are bound by this
transcription factor100,102, which suggests that repression of inflam-
matory genes may occur via indirect binding of STAT6 to DNA and/
or through the recruitment of transcriptional repressors such as
HDAC3, PPAR-γ and components of nuclear receptor–co-repressor
complexes100,104,105. Notably, binding of STAT1, C/EBPα, C/EBPβ or
the NF-κB subunit p65 in response to IFN-γ or LPS is antagonized
by IL-4 at only small sets of enhancers100,102, which indicates that
direct interference with occupancy by inflammatory transcrip-
tion factors is not the main mechanism of repression. In a recip-
rocal fashion, IFN-γ has been reported to suppress IL-4-inducible
genes via interference with chromatin remodeling101,102,106. In human
macrophages, 24 hours of exposure to IFN-γ elicits deposition of
the repressive chromatin mark H3K27me3 at regulatory elements
of genes encoding molecules with immunomodulatory functions,
such as MERTK, PPARG and RANK106. This effect is mediated by
EZH2, the catalytic subunit of the polycomb repressive complex
PRC2, and prevents subsequent induction of those genes by IL-4
and glucocorticoids106. Prolonged (24-hour) stimulation of human
macrophages with IFN-γ disassembles the enhancers of genes
encoding molecules with reparative and immunosuppressive func-
tions that were bound by transcription factors of the MAF family in
unstimulated macrophages101, consistent with the observation that
enhancers antagonized by IL-4 show enrichment for MAFB binding
Nature Immunology | www.nature.com/natureimmunology
FOCUS | Review
| Article
FOCUShttps://doi.org/10.1038/s41590-019-0399-9
Review Article
Box 1 | Known unknowns of adaptation and memory in the
immune system
To what extent are persistent adaptive phenotypes driven by sus-
tained stimulation rather than by bona fide transcriptional and/
or epigenetic memory?
What are the discriminating properties of stimuli that induce
rapidly reversible adaptive responses versus those that induce
sustained adaptive responses?
What is the specific contribution of transcription factor–
mediated memory to the adapted state, relative to that of
chromatin mark–mediated memory?
What are the mechanisms that link signaling, metabolic
and chromatin reprogramming in the establishment and
maintenance of the adapted state?
sites in co-stimulated mouse macrophages102. Such data indicate
that MAF transcription factors might contribute to the ability of
inflammatory stimuli to antagonize gene expression induced by
immunomodulatory stimuli. The co-existence of inflammatory
signals and anti-inflammatory signals that antagonistically con-
trol macrophage activities is common in vivo, in contexts such as
co-infection with helminths (which activate type 2 immunity) and
bacteria or viruses (which activate type 1 immune responses)107–109,
cancer110 and tissue repair111. Furthermore, while the examples
reported here have focused mostly on macrophages, many other
immune-cell and non–immune-cell types can be subjected to simi-
lar regulation. For example, commensal-specific skin T lymphocytes
co-express RORγt and GATA-3, two transcription factors that gov-
ern the antagonistic type 17 helper T cell–differentiation program
and type 2 helper T cell–differentiation program, respectively112. A
hybrid chromatin landscape in these cells supports the concomi-
tant transcription of genes such as Il17a, Il23r and Ccr6 (type 17)
as well as Il5, Il3 and Ccr8 (type 2), which allows plastic adaptations
of commensal-specific T cells between homeostatic conditions and
tissue-injury conditions112.
Conclusions
Data obtained in recent years have contributed to appreciation for
the relevance of adaptive processes to immunological homeostasis
and to the tuning of responses to stimuli. Long-term alterations
in gene expression and associated functional outcomes in innate
immune cells exposed to particular stimuli have often been inter-
preted as evidence of memory of the adapted state. Here we propose
a revision of the current ideas in the field, based on the concept that
bona fide memory in biological systems should possess two essential
properties: specificity of the response, and the presence of dedicated
maintenance mechanisms that make memory a stable and potentially
irreversible state. In B cell and T cell memory, specificity is provided
by antigen receptors, and maintenance of the memory state is linked
to DNA hypomethylation of the promoters and enhancers of genes
encoding critical effector molecules, which maximizes their induc-
tion at re-stimulation. Passive DNA demethylation is generated by
the clonal expansion that occurs after naive lymphocytes encounter
antigen and is stable over decades. Conversely, the enhanced second-
ary responses in primed (or trained) innate immune cells tend to
occur after exposure to unrelated stimuli, and the enhanced activity
state reverts to the basal state in a (relatively) short time. We suggest
that such responses are better described as adaptive phenotypes that
persist because the underlying molecular processes are not efficiently
and rapidly reverted except if cells are exposed to a novel stimulus,
as shown by the swift tissue-instructed transcriptional changes
observed in macrophages in transfer experiments in  vivo28,29. The
persistence of the adapted state is in many cases an advantageous
Review Article | FOCUS
https://doi.org/10.1038/s41590-019-0399-9
and desirable outcome of the primary stimulation, as exemplified
by the reduced production of inflammatory and potentially noxious
cytokines by tolerized macrophages generated by sustained exposure
to large amounts of activating stimuli. In other cases, persistence of
the adaptive phenotype might anticipate future likely events, such
as re-infection with similar microbes, to which the ‘trained’ innate
immune cells would respond faster and more efficiently.
Within the conceptual framework outlined above, many mecha-
nistic details remain to be addressed. The understanding of how
immune cells integrate complex environmental signals into spe-
cific functional states is incomplete, and understanding of how
some inputs generate sustained adaptations is even less complete
(Box  1). A critical issue that is still to be solved at the molecular
level is the role of sustained stimulation relative to that of bona fide
transcriptional or epigenetic memory mechanisms in the genera-
tion of persistent adaptive phenotypes. It remains possible that the
persistence of some stimuli, possibly because of the absence of dedi-
cated degradation systems, may contribute to sustained phenotypes
in the innate immune system, at least in some settings. From an
evolutionary perspective, we are tempted to speculate that adapta-
tions to constant exposure to signals such as β-glucans and other
bioactive polysaccharides that are widely present in food, parasites
or commensals113 may have contributed to the optimization of
immune responses through low-level priming of innate immune
cells. Delineating the rules of signal integration will surely lead to
better understanding of immunological adaptations in vivo, when
cells are concomitantly or sequentially exposed to a changing envi-
ronment milieu. Such knowledge may in turn translate into novel
strategies for modulating immune responses in conditions such as
chronic infection, autoimmunity or cancer.
Received: 8 March 2019; Accepted: 12 April 2019;
Published: xx xx xxxx
References
1. Wilson, C. B., Rowell, E. & Sekimata, M. Epigenetic control of T-helper-cell
differentiation. Nat. Rev. Immunol. 9, 91–105 (2009).
2. Monticelli, S. DNA (hydroxy)methylation in T helper lymphocytes. Trends
Biochem. Sci. https://doi.org/10.1016/j.tibs.2019.01.009 (2019).
3. Netea, M. G. et al. Trained immunity: A program of innate immune
memory in health and disease. Science 352, aaf1098 (2016).
4. Monticelli, S. & Natoli, G. Short-term memory of danger signals and
environmental stimuli in immune cells. Nat. Immunol. 14, 777–784 (2013).
5. Seeley, J. J. & Ghosh, S. Molecular mechanisms of innate memory and
tolerance to LPS. J. Leukoc. Biol. 101, 107–119 (2017).
6. Sun, J. C., Ugolini, S. & Vivier, E. Immunological memory within the innate
immune system. EMBO J. 33, 1295–1303 (2014).
7. O’Sullivan, T. E., Sun, J. C. & Lanier, L. L. Natural killer cell memory.
Immunity 43, 634–645 (2015).
8. Sun, J. C., Beilke, J. N. & Lanier, L. L. Adaptive immune features of natural
killer cells. Nature 457, 557–561 (2009).
9. DuPage, M. & Bluestone, J. A. Harnessing the plasticity of CD4+ T cells to
treat immune-mediated disease. Nat. Rev. Immunol. 16, 149–163 (2016).
10. Sen, D. R. et al. The epigenetic landscape of T cell exhaustion. Science 354,
1165–1169 (2016).
11. Pauken, K. E. et al. Epigenetic stability of exhausted T cells limits durability
of reinvigoration by PD-1 blockade. Science 354, 1160–1165 (2016).
12. Matzinger, P. Friendly and dangerous signals: is the tissue in control? Nat.
Immunol. 8, 11–13 (2007).
13. Varol, C., Mildner, A. & Jung, S. Macrophages: development and tissue
specialization. Annu. Rev. Immunol. 33, 643–675 (2015).
14. Monticelli, S. & Natoli, G. Transcriptional determination and functional
specificity of myeloid cells: making sense of diversity. Nat. Rev. Immunol.
17, 595–607 (2017).
15. Glass, C. K. & Natoli, G. Molecular control of activation and priming in
macrophages. Nat. Immunol. 17, 26–33 (2016).
16. Ghisletti, S. et al. Identification and characterization of enhancers
controlling the inflammatory gene expression program in macrophages.
Immunity 32, 317–328 (2010).
17. Heinz, S. et al. Simple combinations of lineage-determining transcription
factors prime cis-regulatory elements required for macrophage and B cell
identities. Mol. Cell 38, 576–589 (2010).
NATure Immunology
18. Mancino, A. et al. A dual cis-regulatory code links IRF8 to constitutive and
inducible gene expression in macrophages. Genes Dev. 29, 394–408 (2015).
19. Garber, M. et al. A high-throughput chromatin immunoprecipitation
approach reveals principles of dynamic gene regulation in mammals. Mol.
Cell 47, 810–822 (2012).
20. Ostuni, R. et al. Latent enhancers activated by stimulation in differentiated
cells. Cell 152, 157–171 (2013).
21. Kaikkonen, M. U. et al. Remodeling of the enhancer landscape during
macrophage activation is coupled to enhancer transcription. Mol. Cell 51,
310–325 (2013).
22. Okabe, Y. & Medzhitov, R. Tissue-specific signals control reversible
program of localization and functional polarization of macrophages. Cell
157, 832–844 (2014).
23. Gautier, E. L. et al. Gata6 regulates aspartoacylase expression in resident
peritoneal macrophages and controls their survival. J. Exp. Med. 211,
1525–1531 (2014).
24. Rosas, M. et al. The transcription factor Gata6 links tissue macrophage
phenotype and proliferative renewal. Science 344, 645–648 (2014).
25. Haldar, M. et al. Heme-mediated SPI-C induction promotes monocyte
differentiation into iron-recycling macrophages. Cell 156, 1223–1234 (2014).
26. Scott, C. L. et al. The transcription factor ZEB2 is required to maintain the
tissue-specific identities of macrophages. Immunity 49, 312–325.e315 (2018).
27. Mass, E. et al. Specification of tissue-resident macrophages during
organogenesis. Science 353, aaf4238 (2016).
28. Lavin, Y. et al. Tissue-resident macrophage enhancer landscapes are shaped
by the local microenvironment. Cell 159, 1312–1326 (2014).
29. Gosselin, D. et al. Environment drives selection and function of enhancers
controlling tissue-specific macrophage identities. Cell 159, 1327–1340 (2014).
30. Deczkowska, A. et al. Mef2C restrains microglial inflammatory response
and is lost in brain ageing in an IFN-I-dependent manner. Nat. Commun. 8,
717 (2017).
31. Matcovitch-Natan, O. et al. Microglia development follows a stepwise
program to regulate brain homeostasis. Science 353, aad8670 (2016).
32. Saijo, K. et al. A Nurr1/CoREST pathway in microglia and astrocytes
protects dopaminergic neurons from inflammation-induced death. Cell 137,
47–59 (2009).
33. Kayama, H. et al. Heme ameliorates dextran sodium sulfate-induced colitis
through providing intestinal macrophages with noninflammatory profiles.
Proc. Natl Acad. Sci. USA 115, 8418–8423 (2018).
34. Casanova-Acebes, M. et al. Neutrophils instruct homeostatic and
pathological states in naive tissues. J. Exp. Med. 215, 2778–2795 (2018).
35. Ng, L. G., Ostuni, R. & Hidalgo, A. Heterogeneity of neutrophils. Nat. Rev.
Immunol. 19, 255–265 (2019).
36. Casanova-Acebes, M. et al. Rhythmic modulation of the hematopoietic
niche through neutrophil clearance. Cell 153, 1025–1035 (2013).
37. Puga, I. et al. B cell-helper neutrophils stimulate the diversification and
production of immunoglobulin in the marginal zone of the spleen. Nat.
Immunol. 13, 170–180 (2011).
38. Mackay, L. K. & Kallies, A. Transcriptional regulation of tissue-resident
lymphocytes. Trends Immunol. 38, 94–103 (2017).
39. Ditadi, A., Sturgeon, C. M. & Keller, G. A view of human haematopoietic
development from the Petri dish. Nat. Rev. Mol. Cell Biol. 18, 56–67 (2017).
40. Ginhoux, F. & Guilliams, M. Tissue-resident macrophage ontogeny and
homeostasis. Immunity 44, 439–449 (2016).
41. Gentek, R. et al. Hemogenic endothelial fate mapping reveals dual
developmental origin of mast cells. Immunity 48, 1160–1171.e1165 (2018).
42. Li, Z. et al. Adult connective tissue-resident mast cells originate from late
erythro-myeloid progenitors. Immunity 49, 640–653.e645 (2018).
43. Smith, F. L. & Baumgarth, N. B-1 cell responses to infections. Curr. Opin.
Immunol. 57, 23–31 (2019).
44. Gentek, R. et al. Epidermal γ δ T cells originate from yolk sac
hematopoiesis and clonally self-renew in the adult. J. Exp. Med. 215,
2994–3005 (2018).
45. Smith, N. L. et al. Developmental origin governs CD8 + T cell fate decisions
during infection. Cell 174, 117–130.e114 (2018).
46. Shemer, A. et al. Engrafted parenchymal brain macrophages differ from
microglia in transcriptome, chromatin landscape and response to challenge.
Nat. Commun. 9, 5206 (2018).
47. Cronk, J. C. et al. Peripherally derived macrophages can engraft the brain
independent of irradiation and maintain an identity distinct from
microglia. J. Exp. Med. 215, 1627–1647 (2018).
48. Zhu, Y. et al. Tissue-resident macrophages in pancreatic ductal
adenocarcinoma originate from embryonic hematopoiesis and promote
tumor progression. Immunity 47, 597 (2017).
49. Bowman, R. L. et al. Macrophage ontogeny underlies differences in
tumor-specific education in brain malignancies. Cell Rep. 17,
2445–2459 (2016).
50. Pope, S. D. & Medzhitov, R. Emerging Principles of Gene Expression
Programs and Their Regulation. Mol. Cell 71, 389–397 (2018).
Nature Immunology | www.nature.com/natureimmunology

NATure Immunology
51. Smale, S. T., Tarakhovsky, A. & Natoli, G. Chromatin contributions
to the regulation of innate immunity. Annu. Rev. Immunol. 32,
489–511 (2014).
52. Foster, S. L., Hargreaves, D. C. & Medzhitov, R. Gene-specific control of
inflammation by TLR-induced chromatin modifications. Nature 447,
972–978 (2007).
53. Kamada, R. et al. Interferon stimulation creates chromatin marks and
establishes transcriptional memory. Proc. Natl Acad. Sci. USA 115,
E9162–E9171 (2018).
54. Daniel, B. et al. The nuclear receptor PPARγ controls progressive
macrophage polarization as a ligand-insensitive epigenomic ratchet of
transcriptional memory. Immunity 49, 615–626.e616 (2018).
55. Qiao, Y. et al. Synergistic activation of inflammatory cytokine genes by
interferon-γ-induced chromatin remodeling and toll-like receptor signaling.
Immunity 39, 454–469 (2013).
56. Cheng, S. C. et al. mTOR- and HIF-1α-mediated aerobic glycolysis as
metabolic basis for trained immunity. Science 345, 1250684 (2014).
57. Saeed, S. et al. Epigenetic programming of monocyte-to-macrophage
differentiation and trained innate immunity. Science 345, 1251086 (2014).
58. Wendeln, A. C. et al. Innate immune memory in the brain shapes
neurological disease hallmarks. Nature 556, 332–338 (2018).
59. Naik, S. et al. Inflammatory memory sensitizes skin epithelial stem cells to
tissue damage. Nature 550, 475–480 (2017).
60. Ordovas-Montanes, J. et al. Allergic inflammatory memory in human
respiratory epithelial progenitor cells. Nature 560, 649–654 (2018).
61. McLane, L.M., Abdel-Hakeem, M.S. & Wherry, E.J. CD8 T cell exhaustion
during chronic viral infection and cancer. Annu. Rev. Immunol. 37,
457–495 (2019).
62. Hotchkiss, R. S., Monneret, G. & Payen, D. Sepsis-induced
immunosuppression: from cellular dysfunctions to immunotherapy.
Nat. Rev. Immunol. 13, 862–874 (2013).
63. Chen, J. & Ivashkiv, L. B. IFN-γ abrogates endotoxin tolerance by
facilitating Toll-like receptor-induced chromatin remodeling. Proc. Natl
Acad. Sci. USA 107, 19438–19443 (2010).
64. Novakovic, B. et al. β-Glucan reverses the epigenetic state of LPS-induced
immunological tolerance. Cell 167, 1354–1368.e1314 (2016).
65. Park, S. H. et al. Type I interferons and the cytokine TNF cooperatively
reprogram the macrophage epigenome to promote inflammatory activation.
Nat. Immunol. 18, 1104–1116 (2017).
66. Burrill, D. R. & Silver, P. A. Making cellular memories. Cell 140,
13–18 (2010).
67. Ptashne, M. Regulation of transcription: from lambda to eukaryotes. Trends
Biochem. Sci. 30, 275–279 (2005).
68. Seeley, J. J. et al. Induction of innate immune memory via microRNA
targeting of chromatin remodelling factors. Nature 559, 114–119 (2018).
69. Ramirez-Carrozzi, V. R. et al. A unifying model for the selective regulation
of inducible transcription by CpG islands and nucleosome remodeling. Cell
138, 114–128 (2009).
70. Toyama, B. H. et al. Identification of long-lived proteins reveals exceptional
stability of essential cellular structures. Cell 154, 971–982 (2013).
71. Ghoneim, H. E. et al. De novo epigenetic programs inhibit PD-1
blockade-mediated T cell rejuvenation. Cell 170, 142–157.e119 (2017).
72. Mitroulis, I. et al. Modulation of myelopoiesis progenitors is an integral
component of trained immunity. Cell 172, 147–161.e112 (2018).
73. Kouzarides, T. Chromatin modifications and their function. Cell 128,
693–705 (2007).
74. Deal, R. B., Henikoff, J. G. & Henikoff, S. Genome-wide kinetics of
nucleosome turnover determined by metabolic labeling of histones. Science
328, 1161–1164 (2010).
75. Dion, M. F. et al. Dynamics of replication-independent histone turnover in
budding yeast. Science 315, 1405–1408 (2007).
76. Kraushaar, D. C. et al. Genome-wide incorporation dynamics reveal
distinct categories of turnover for the histone variant H3.3. Genome Biol.
14, R121 (2013).
77. Margueron, R. et al. Role of the polycomb protein EED in the propagation
of repressive histone marks. Nature 461, 762–767 (2009).
78. Quintin, J. et al. Candida albicans infection affords protection against
reinfection via functional reprogramming of monocytes. Cell Host Microbe
12, 223–232 (2012).
79. Stone, B.A. Chemistry of β-glucans. in Chemistry, Biochemistry, and Biology
of 1→3-β-Glucans and Related Polysaccharides (eds. Bacic, A. et al.) Ch. 2.1
(Academic Press, 2009).
80. Williams, D. L. Overview of (1→3)-β-D-glucan immunobiology. Mediators
Inflamm. 6, 247–250 (1997).
81. Ariizumi, K. et al. Identification of a novel, dendritic cell-associated
molecule, dectin-1, by subtractive cDNA cloning. J. Biol. Chem. 275,
20157–20167 (2000).
82. Brown, G. D. & Gordon, S. Immune recognition. A new receptor for
β-glucans. Nature 413, 36–37 (2001).
Nature Immunology | www.nature.com/natureimmunology
FOCUS | Review
| Article
FOCUShttps://doi.org/10.1038/s41590-019-0399-9
Review Article
83. Palma, A. S. et al. Ligands for the β-glucan receptor, Dectin-1, assigned
using “designer” microarrays of oligosaccharide probes (neoglycolipids)
generated from glucan polysaccharides. J. Biol. Chem. 281,
5771–5779 (2006).
84. Nono, I., Ohno, N., Masuda, A., Oikawa, S. & Yadomae, T. Oxidative
degradation of an antitumor (1→3)-β-D-glucan, grifolan. J. Pharmacobiodyn.
14, 9–19 (1991).
85. Ozment, T. R., Goldman, M. P., Kalbfleisch, J. H. & Williams, D. L. Soluble
glucan is internalized and trafficked to the Golgi apparatus in macrophages
via a clathrin-mediated, lipid raft-regulated mechanism. J. Pharmacol. Exp.
Ther. 342, 808–815 (2012).
86. Miura, T., Ohno, N., Miura, N. N., Shimada, S. & Yadomae, T. Inactivation
of a particle β-glucan by proteins in plasma and serum. Biol. Pharm. Bull.
20, 1103–1107 (1997).
87. Miura, T. et al. Inactivation of (1→3)-β-D-glucan in mice. Biol. Pharm.
Bull. 18, 1797–1801 (1995).
88. Chen, K. et al. Super-low dose endotoxin pre-conditioning exacerbates
sepsis mortality. EBioMedicine 2, 324–333 (2015).
89. Lira, F. S. et al. Endotoxin levels correlate positively with a sedentary
lifestyle and negatively with highly trained subjects. Lipids Health Dis. 9,
82 (2010).
90. Szeto, C. C. et al. Endotoxemia is related to systemic inflammation and
atherosclerosis in peritoneal dialysis patients. Clin. J. Am. Soc. Nephrol. 3,
431–436 (2008).
91. Wiedermann, C. J. et al. Association of endotoxemia with carotid
atherosclerosis and cardiovascular disease: prospective results from the
Bruneck Study. J. Am. Coll. Cardiol. 34, 1975–1981 (1999).
92. Olsen, A. W., Brandt, L., Agger, E. M., van Pinxteren, L. A. & Andersen, P.
The influence of remaining live BCG organisms in vaccinated mice on
the maintenance of immunity to tuberculosis. Scand. J. Immunol. 60,
273–277 (2004).
93. Kaveh, D. A., Garcia-Pelayo, M. C. & Hogarth, P. J. Persistent BCG bacilli
perpetuate CD4 T effector memory and optimal protection against
tuberculosis. Vaccine 32, 6911–6918 (2014).
94. Queiroz, A. & Riley, L. W. Bacterial immunostat: Mycobacterium
tuberculosis lipids and their role in the host immune response. Rev. Soc.
Bras. Med. Trop. 50, 9–18 (2017).
95. Xue, J. et al. Transcriptome-based network analysis reveals a spectrum
model of human macrophage activation. Immunity 40, 274–288 (2014).
96. Murray, P. J. et al. Macrophage activation and polarization: nomenclature
and experimental guidelines. Immunity 41, 14–20 (2014).
97. Schroder, K. & Tschopp, J. The inflammasomes. Cell 140, 821–832 (2010).
98. Zanoni, I. et al. An endogenous caspase-11 ligand elicits interleukin-1
release from living dendritic cells. Science 352, 1232–1236 (2016).
99. Saccani, S., Pantano, S. & Natoli, G. Two waves of nuclear factor kB
recruitment to target promoters. J. Exp. Med. 193, 1351–1359 (2001).
100. Czimmerer, Z. et al. The transcription factor STAT6 mediates direct
repression of inflammatory enhancers and limits activation of alternatively
polarized macrophages. Immunity 48, 75–90.e76 (2018).
101. Kang, K. et al. Interferon-γ represses m2 gene expression in human
macrophages by disassembling enhancers bound by the transcription factor
MAF. Immunity 47, 235–250.e234 (2017).
102. Piccolo, V. et al. Opposing macrophage polarization programs show
extensive epigenomic and transcriptional cross-talk. Nat. Immunol. 18,
530–540 (2017).
103. Schleicher, U. et al. TNF-mediated restriction of arginase 1 expression in
myeloid cells triggers type 2 NO synthase activity at the site of infection.
Cell Rep. 15, 1062–1075 (2016).
104. Nelson, V. L. et al. PPARγ is a nexus controlling alternative activation of
macrophages via glutamine metabolism. Genes Dev. 32, 1035–1044 (2018).
105. Glass, C. K. & Saijo, K. Nuclear receptor transrepression pathways that
regulate inflammation in macrophages and T cells. Nat. Rev. Immunol. 10,
365–376 (2010).
106. Qiao, Y., Kang, K., Giannopoulou, E., Fang, C. & Ivashkiv, L. B. IFN-γ
induces histone 3 lysine 27 trimethylation in a small subset of promoters to
stably silence gene expression in human macrophages. Cell Rep. 16,
3121–3129 (2016).
107. Ezenwa, V. O. & Jolles, A. E. Epidemiology. Opposite effects of anthelmintic
treatment on microbial infection at individual versus population scales.
Science 347, 175–177 (2015).
108. Osborne, L. C. et al. Coinfection. Virus-helminth coinfection reveals a
microbiota-independent mechanism of immunomodulation. Science 345,
578–582 (2014).
109. Rückerl, D. et al. Macrophage origin limits functional plasticity in
helminth-bacterial co-infection. PLoS Pathog. 13, e1006233 (2017).
110. Kratochvill, F. et al. TNF counterbalances the emergence of M2 tumor
macrophages. Cell Rep. 12, 1902–1914 (2015).
111. Bosurgi, L. et al. Macrophage function in tissue repair and remodeling
requires IL-4 or IL-13 with apoptotic cells. Science 356, 1072–1076 (2017).
Review Article | FOCUS
https://doi.org/10.1038/s41590-019-0399-9 NATure Immunology
112. Harrison, O.J. et al. Commensal-specific T cell plasticity promotes rapid Ministry of Health (GR-2016-02362156), the Associazione Italiana per la Ricerca sul
tissue adaptation to injury. Science 363, eaat6280 (2019). Cancro (AIRC MFAG, 20247), the Cariplo Foundation (2015–0990) and the European
113. Blander, J. M., Longman, R. S., Iliev, I. D., Sonnenberg, G. F. & Artis, D. Union (Infect-ERA 126).
Regulation of inflammation by microbiota interactions with the host.
Nat. Immunol. 18, 851–860 (2017). Competing interests
The authors declare no competing interests.
Acknowledgements
We are grateful to members of the G.N. and R.O. laboratories for scientific discussions
and to S. Monticelli for critical reading of the manuscript. Research on related topics in Additional information
the G.N. laboratory is supported by the European Research Council (ERC Advanced Reprints and permissions information is available at www.nature.com/reprints.
Grant 692789, MEDICI), the European Commission (Consortium grant SYSCYD), the Correspondence should be addressed to G.N. or R.O.
Cariplo Foundation and the Italian Ministry of University and Research (MIUR, grant
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in
METRIC (METabolic Regulation of Inflammatory Cells). Research in the R.O. laboratory
published maps and institutional affiliations.
is supported by grants from the European Research Council (ERC Starting Grant
759532, X-TAM), the Italian Telethon Foundation (SR-Tiget grant award F04), the Italian © Springer Nature America, Inc. 2019

Nature Immunology | www.nature.com/natureimmunology