Trained immunity and host-pathogen interactions

Summary

Infectious diseases are a leading cause of death worldwide with over 8 million fatalities accounted for in 2016. Solicitation of host immune defenses by vaccination is the treatment of choice to prevent these infections. It has long been thought that vaccine immunity was solely mediated by the adaptive immune system. However, over the past decade, numerous studies have shown that innate immune cells can also retain memory of these encounters. This process, called innate immune memory is mediated by metabolic and epigenetic changes that make cells either hyperresponsive (trained immunity) or hyporesponsive (tolerance) to subsequent challenges. In this review, we discuss the concepts of trained immunity and tolerance in the context of host-pathogen interactions.

Introduction

Infectious diseases claimed the lives of over 8 million people in 2017, and over 2.6 million of new infection cases are reported in the United States each year (CDC, 2018; Foreman et al., 2018). There are several reasons behind these significant morbidity and mortality rates. The global rise of antimicrobial resistance continues to thwart efforts to effectively treat many of these infections (Prestinaci, Pezzotti, & Pantosti, 2015) and while some vaccines may exist, their efficacy against all strains of a given pathogen can remain incomplete. There are also social, economic and environmental factors such as population growth, civil turmoil and natural disasters that can lead to dire conditions propitious to infectious diseases development. Other aspects linked to globalization, such as increases in international air traffic and food trade also considerably heighten the risks of fast spread of infectious agents (Drexler, 2010) and the emergence of new viral pathogens such as Ebola, Zika and SARS-CoV-2, for which, treatment and preventive options are limited (Bloom & Cadarette, 2019).

One way to protect individuals from infections is through activation of host immune defenses. The development of routine vaccination is one of the most impactful events of the 20^th century, with major effects on reducing mortality and aiding population growth (Plotkin & Plotkin, 2018). This is best illustrated by prevention of the spread of infectious agents such as polio virus and the global eradication of smallpox (Piot et al., 2019). However, traditional vaccines rely on adaptive immunity that is restricted to specific antigens and has a delayed response (Janeway, Travers, Walport, & Shlomchik, 2001). Over the past decade, studies have revealed that vaccine immunity could also partially rely on innate immune mediators, a concept that has been called trained immunity (Netea, Quintin, & van der Meer, 2011). Understanding the key modulators of this phenomenon, including microbial triggers, cytokines, cells and intracellular factors is essential for the potential development of effective prophylactic or therapeutic treatments. In this review we will describe the concepts of trained immunity and discuss recent data that support its role in host defense with an emphasis on host-pathogen interactions.

Innate immune memory

Immunological memory is defined as “the ability of the immune system to respond more rapidly and effectively to pathogens that have been encountered previously” (Janeway et al., 2001). It was long thought that memory was an exclusive property of the adaptive immune system. However, studies in plants, invertebrates (that both lack adaptive immunity) and, more recently, in mammals, have shown that innate immune cells can also perform this function (Kachroo & Robin, 2013; Kurtz, 2005; Quintin, Cheng, van der Meer, & Netea, 2014). The ability of innate immune cells to remember and respond accordingly to secondary challenges has thus been termed “innate immune memory” (Rodrigues, Brayner, Alves, Dixit, & Barillas-Mury, 2010). The innate immune memory phenotype is highly dependent on the nature and intensity of external stimuli and can result either in enhanced activation, also known as “trained immunity” (Netea et al., 2011), or unresponsiveness, also known as “tolerance” (Bigot et al., 2020). In this context, the term tolerance is distinct from self-tolerance, defined as a lack of immunological response to self-antigens.

Tolerance.

Tolerance is believed to be a protective mechanism established to decrease host susceptibility to tissue damage caused by pathogens or the subsequent immune response against them (Medzhitov, Schneider, & Soares, 2012). The downside of this process is that it can result in immune paralysis, where cells are unable to produce pro-inflammatory mediators and, thus, render the host more susceptible to secondary infection. Lipopolysaccharide (LPS), a component of the cell membrane of Gram-negative bacteria, is a potent stimulator of innate immune responses. Dysregulation of this pathway is linked with excessive inflammation and tissue damage. Endotoxin tolerance occurs after persistent toll-like receptor stimulation by LPS (16 to 48 hours) and, in macrophages, leads to reduced production of tumor necrosis factor (TNF)-α, interleukin (IL)-12, IL-6 and increased production of anti-inflammatory cytokines such as IL-10 and transforming growth factor (TGF)-β upon secondary stimulation (Vergadi, Vaporidi, & Tsatsanis, 2018). Pretreatment with high doses of LPS (between 10 ng/mL and 1 μg/mL) triggers a greater degree of tolerization (Seeley & Ghosh, 2017). This phenotype is also observed in leukocytes of patients suffering from sepsis (Biswas & Lopez-Collazo, 2009). It is interesting to note that, despite their hyporesponsive state, cells that undergo LPS desensitization still preserve some antimicrobial functions such as the production of cathelicidin-related antimicrobial peptides (Camp) and lipocalin 2 (Foster, Hargreaves, & Medzhitov, 2007). Ex-vivo pre-treatment of human monocytes, with exogenous interferon γ (IFN-γ) was shown to prevent the endotoxin tolerization phenotype (Chen & Ivashkiv, 2010). LPS tolerization was also shown to be reversed upon exposure to the fungal polysaccharide, β-glucan, in human monocytes (Novakovic et al., 2016). Immune paralysis has also been described in human and mouse alveolar macrophages (AM) after trauma, sepsis and pneumonia and results in decreased phagocytic abilities, a process mediated by the signal-regulatory protein SIRPα (Roquilly et al., 2020). In addition, tolerance appears to be involved in influenza induced susceptibility to secondary infections with respiratory syncytial virus, Cryptococcus neoformans, Streptococcus pneumoniae and Escherichia coli (Didierlaurent et al., 2008; Roquilly et al., 2017). Several tolerization studies demonstrate that LPS tolerance does not only suppress inflammatory damage but also plays a role in innate immune memory by protecting the host from subsequent infections. Pre-treatment of mice with LPS was shown to reduce pathogen spread and /or increase survival in the context of polymicrobial sepsis, Cryptococcus neoformans, Salmonella enterica, Staphylococcus aureus and group B streptococcus infections (Seeley & Ghosh, 2017). This protection phenotype is thought to be the result of changes in expression of non-tolerizable genes. One mechanism described in this context is the effect of LPS treatment on the stress-response transcription factor ATF7. ATF7 acts as suppressor of a group of genes encoding factors involved in innate immunity. Upon LPS treatment, ATF7 is phosphorylated and released from the chromatin, allowing expression of target genes and resistance to subsequent infections (Yoshida et al., 2015).

Trained immunity.

One of the first observations that led to the discovery of trained immunity was that vaccination with bacillus Calmette-Guérin (BCG), an attenuated strain of Mycobacterium bovis utilized to elicit protective immunity against tuberculosis (TB), induced a significant decrease in mortality that could not be explained by protection against TB alone. Instead, its protective effect seemed to extend to other conditions such as sepsis and respiratory infections (Garly et al., 2003; Roth et al., 2005). Two possible mechanisms could explain the nonspecific effects of this vaccination. The first one is the elicitation of cross-protection by heterologous T cell memory responses (Welsh & Selin, 2002). However, this mechanism relies on a functional adaptive immune system and needs at least 14 days to develop and is, therefore, unlikely to explain the observed protective effects of BCG immunization in newborn infants (Biering-Sørensen et al., 2012). The second proposed mechanism is the epigenetic reprogramming of innate immune cells, a phenomenon that confers sustained modifications in transcription programs and leads to changes in cell physiology and responses to subsequent infections, also called trained immunity (Netea, Domínguez-Andrés, et al., 2020). Furthermore, BCG vaccination has been shown to confer T- and B-lymphocyte-independent protection in the context of disseminated candidiasis in severe combined immunodeficiency (SCID) mice, suggesting that trained immunity is, at least, partially responsible for these non-specific protective effects (Kleinnijenhuis et al., 2012). Interestingly, in this context, BCG vaccination was found to specifically enhance the functions of circulating monocytes to produce cytokines such as TNF-α and IL-1β via epigenetic remodeling, such as methylation of histone H3 at lysine 4 (H3K4) (Kleinnijenhuis et al., 2012). Similar observations (increased TNF-α production upon secondary challenge with LPS and epigenetic remodeling) were detected in monocytes stimulated with β-glucan (Cheng et al., 2014).

BCG and β-glucan are the most studied stimuli of trained immunity and can elicit protection against secondary infections with a range of pathogens. For example, in humans, BCG vaccination affords protection against yellow fever virus infection (R. J. Arts et al., 2018) and in mice β-glucan inoculation can protect against Staphylococcus aureus, Listeria monocytogenes and Pseudomonas aeruginosa infections (Ciarlo et al., 2019). However, other agents can also perform this function. For instance, the bacterial component flagellin has been shown to protect against Aspergillus fumigatus infections in human bronchial epithelial cells (hBEC) (Bigot et al., 2020). Flagellin also has a protective effect against Salmonella typhimurium and rotavirus infections in mice but whether this solely relies on trained immunity needs to be further investigated (Vijay-Kumar et al., 2008; B. Zhang et al., 2014). Furthermore, other pathogens can also induce trained immunity. For example, S. pneumoniae infection was shown to cause long-term phenotypic changes, including decreased expression of SiglecF and increased expression of MHCII, CD64 and CD11c in murine alveolar macrophages as well as metabolomic changes, with increased phosphocreatine and creatine production in trained cells that resulted in enhanced responsiveness and clearance of secondary bacterial infections (Guillon et al., 2020). Alveolar macrophages from S. pneumoniae-colonized individuals also displayed functional changes with increased opsonophagocytic activity and nonspecific protection against subsequent ex vivo challenges with Streptococcus pyogenes and S. aureus compared to AM from individuals not colonized with S. pneumoniae (Mitsi et al., 2020). In models of planarian and mouse skin infections, priming with S. aureus was also shown to confer protection against reinfection (Chan et al., 2018; Torre et al., 2017; Wong Fok Lung et al., 2020). Finally, non-microbial factors can also play a role in trained immunity. For example, mevalonate (metabolite of the cholesterol biosynthesis pathway), oxidized low-density lipoprotein (oxLDL) and western diet have all been shown to directly induce transcriptomic and epigenomic remodeling of innate immune cells (Bekkering et al., 2018; Christ et al., 2018; Sohrabi et al., 2019).

Adaptive immune memory can last for years, sometimes even a lifetime but the extent of innate immune memory persistence remains to be fully elucidated. So far, studies have shown this phenotype can last for as long as three months to one year (Netea, Domínguez-Andrés, et al., 2020). The fact that LPS desensitization can be reversed by other stimuli (Novakovic et al., 2016) suggests that trained immunity may be reversible, but the induced reversal of trained immunity has not been a subject of investigation to-date.

Epigenetics, immunometabolism and trained immunity

Trained immunity is mediated via epigenetic remodeling of transcriptional pathways. Epigenetics is defined as “the structural adaptation of chromosomal regions so as to register, signal or perpetuate altered activity states” (Bird, 2007), which means that gene expression is modified without altering the genetic code itself. In trained immunity, these events occur through histone acetylation and methylation events that make DNA more accessible to transcription factors at the promoters and enhancers of proinflammatory genes such as TNFA and IL6. Increased levels of trimethylation at the 4^th lysine residue of histone 3 (H3K4me3) have been observed in human monocytes after β-glucan stimulation (Quintin et al., 2012). Exposure of macrophages to the same compound led to modifications in H3K27ac as well as increased numbers of H3K4me1and H3K4me3 marks at enhancers and promoters of genes involved in oxidative reduction, metabolism and lysosome function. This resulted in transcriptionally active chromatin. β-glucan exposure led to decreased marks on genes associated with cytokine and chemokine responses (Novakovic et al., 2016).. Finally, genome-wide epigenetic reprogramming has been detected in human monocytes after in vivo training by BCG vaccination (R. J. Arts et al., 2018). Global changes in H3K27Ac were observed in pathways such as phosphatidylinositol 3-kinase (PI3K), epidermal growth factor receptor (EGFR) and vascular endothelial growth factor (VEGF). These histone marks appear to facilitate quicker and more robust gene expression upon secondary challenge with similar or unrelated stimuli, responding with increased innate (IL-1β, TNF, IL-6) and Th2/T17 cytokines and chemokines. Although epigenomic patterns have been observed in trained cells, their specific roles in innate memory formation are largely unknown (Q. Zhang & Cao, 2019).

There is a strong interplay between epigenetic remodeling and metabolism. The activity of chromatin modifying enzymes is dependent on metabolites such as acetyl-CoA and NAD⁺ that participate in many biochemical reactions in protein, carbohydrate and lipid metabolisms (Etchegaray & Mostoslavsky, 2016). Therefore, epigenetic modifications are directly influenced by the nutritional status of the cell and different stimuli have distinct effects on immunometabolism programs. For example, β-glucan stimulation induces a shift from oxidative phosphorylation (OxPhos) to glycolysis regulated via the mTOR-HIF-1α pathway (Cheng et al., 2014). This stimulation is also known to result in inhibition of Sirtuin1, a histone deacetylase gene (Cheng et al., 2014). Despite the reduction in OxPhos, the tricarboxylic acid (TCA) cycle is not completely shut down and some of its compounds, including citrate, succinate and fumarate are found at higher levels than in non-stimulated cells. Several roles have been attributed to these metabolites: citrate contributes to histone acetylation at the site of glycolytic enzymes genes such as hexokinase 2, phosphofructokinase and lactate dehydrogenase. Succinate and fumarate can inhibit histone and DNA demethylases such as lysine demethylases of the JmjC family and stabilize HIF-1α, thereby contributing to glycolysis (R. J. W. Arts, Joosten, & Netea, 2016). oxLDL stimulation has also been linked with increased glycolysis (Sohrabi et al., 2019). Glutaminolysis and the cholesterol synthesis pathway, in particular the metabolite mevalonate, are also both required for β-glucaninduced trained immunity (R. J. Arts et al., 2016; Bekkering et al., 2018).

BCG stimulation leads to increases in both glycolysis and OxPhos. Inhibition experiments revealed that glycolytic rate and lactate production were more important for induction of trained immunity than OxPhos. Open chromatin and closed chromatin marks (H3K4me3 and H3K9me3) were found to be increased and decreased, respectively, at the promoters of enzymes involved in glucose and glutamine metabolisms suggesting that increased activation of both pathways is involved in this process (R. J. Arts et al., 2016).

LPS stimulation has been shown to upregulate glycolysis and fatty acid synthesis, while suppressing OxPhos. The suppressive effect on OxPhos appears to be specific for higher concentrations of LPS, which result in increased NAD⁺/NADH ratios that activate the histone deacetylases sirtuin 1 and 6 and leads to a switch from proinflammatory state with high glycolysis and OxPhos to a more anti-inflammatory state with increased fatty acid oxidation (Liu, Vachharajani, Yoza, & McCall, 2012; Penkov, Mitroulis, Hajishengallis, & Chavakis, 2019). The itaconate synthesis pathway appears to be the central regulatory node linking immune tolerance and trained immunity. LPS-induced production of itaconate promotes tolerance in human monocytes, while β-glucan counteracts this effect by inhibiting the expression of immune-responsive gene 1 (IRG1), the enzyme that regulates itaconate synthesis (Domínguez-Andrés et al., 2019).

Trained immunity across cell types and tissues

As mentioned in the previous paragraphs, monocytes and macrophages have the ability to undergo epigenetic and functional reprogramming. This reprogramming leads to enhanced proinflammatory cytokine production and protection upon secondary challenge, and has been demonstrated in mouse and human models of infection (R. J. Arts et al., 2018; Cheng et al., 2014; Kleinnijenhuis et al., 2012; Quintin et al., 2012). Interestingly, these trained immunity characteristics appear to be shared by other cell types including natural killer (NK) cells, dendritic cells (DCs), epithelial cells and stem cells.

The ability of bone marrow progenitors to be trained following immunization could explain the long-lasting effects of trained immunity despite the short lifespan of mature myeloid cells in circulation. Hematopoietic stem and progenitor cell (HSPC) training by BCG and β-glucan stimulation have been described and result in increased myelopoiesis in mice (Kaufmann et al., 2018; Mitroulis et al., 2018). These processes appear to be dependent on IFN-γ or IL-1β, respectively. This is also observed in humans, with BCG vaccination imprinting a persistent myeloid transcriptomic bias on HSPCs. This is regulated by the hepatic nuclear factors HNF1a and HNF1b and results in persistent training of CD14⁺ monocytes (Cirovic et al., 2020).

Due to their frequent exposure to microbes, mucosal surfaces such as the airways and the skin are suitable sites for trained immunity. In the lung, influenza A stimulation was shown to induce training of monocyte-derived AM resulting in increased production of IL-6 by these cells and antibacterial protection against subsequent S. pneumoniae infection one month after initial flu challenge (Aegerter et al., 2020). This contrasts with other co-infection studies that see enhanced susceptibility to bacterial infection at earlier times for re-challenge (McCullers & Rehg; Shahangian et al., 2009; Sharma-Chawla et al., 2016; Siegel, Roche, & Weiser, 2014). This possibly relates to the stage of infection, acute vs resolved. In contrast, influenza A stimulation had no major effect on the functional, transcriptional or chromatin profiles of tissue resident AM, with only 60 genes differentially expressed between cells from naïve and post-influenza lungs. This could suggest that intrinsic cell properties play a major role in trained immunity (Aegerter et al., 2020). However, another study showed that, after adenoviral infection, memory AM could develop independently of recruited monocytes and be trained against S. pneumoniae secondary infection. These trained cells displayed an increased propensity to produce MIP-2 and KC, compared to naïve AM, upon ex-vivo stimulation with S. pneumoniae. This protection, however, required initial priming by CD8 T cells, via IFN-γ production (Yao et al., 2018). Trained AM to S. pneumoniae alone show increases in SiglecF and decreased MHCII along with increased in metabolites such as phosphocreatine. AM trained to S. pneumoniae had enhanced signaling upon re-infection, with increases in genes associated with immune signaling and energy production (Guillon et al., 2020). Therefore, different training agents could have distinct effects on specific cells types.

In the skin, epithelial stem cells have also been shown to undergo IL-1β-dependent training leading to improved wound-healing, without the contribution of other cell types (Naik et al., 2017). In the context of repeated S. aureus skin infections in mice, trained immunity was shown to be primarily systemic. In this model, animals were primed with a single subcutaneous injection on one flank. For the secondary challenge, 6 weeks later, animals received subcutaneous injections on both flanks. Primed animals displayed improved wound-healing compared to naïve animals on both their naïve and primed flanks, suggesting systemic protection (Chan et al., 2017). While improved protection was observed on the previously infected flank, suggesting an element of localized protection as well. Protection in the skin was correlated with enhanced production IL-6, IL-17, IFN-γ and RANTES, and the increased presence of Th1, DC and macrophage populations. Macrophages and myeloid precursors from the bone marrow could participate in this response as bone marrow derived macrophages from primed and naïve mice showed differential responses and phagocytosis to subsequent ex-vivo challenges (Chan et al., 2018).

NK cells can also participate in trained immunity. Murine NK cells preactivated with IL-12 and IL-18 were shown to produce abundant amounts of IFN-γ upon restimulation with cytokines or engagement/activationof the NK cell receptors Ly49H and NK1.1. This memory response was shown to be independent of proliferation and transmitted to daughter cells (Cooper et al., 2009). In the context of murine cytomegalovirus infection, memory NK cells were detected up to 70 days post infection and also displayed a heightened IFN-γ response compared to naïve cells (Sun, Beilke, & Lanier, 2009). Furthermore, NK cells of human volunteers vaccinated with BCG showed an increase in IL-1β and IL-6 production upon secondary ex-vivo challenges with Mycobacterium tuberculosis and Staphylococcus aureus (Kleinnijenhuis et al., 2014). Finally, BCG-induced trained immunity against disseminated Candida albicans infection was also found to partially rely on NK cells in mice (Kleinnijenhuis et al., 2014).

A recent study has shown that DCs isolated from mice exposed to Cryptococcus neoformans displayed a memory phenotype with an increased production of IFN-γ and TNF-α upon ex-vivo secondary challenge with the same pathogen 70 days after primary infection (Hole et al., 2019). These cytokine production patterns were greatly reduced in the presence of the histone methyltransferase inhibitor MTA, suggesting that epigenetic remodeling plays a role in this process (Hole et al., 2019). Overall, a wide array of mature innate immune cells can be trained across multiple organs, with different stimuli. It is of particular interest that stem cells can also respond to training stimuli and pass on their memory phenotype to their mature progeny. This aspect of trained immunity could be an advantageous target for the development of long-lasting preventive or therapeutic treatments.

Bacteria and epigenetic remodeling.

While some microorganisms passively induce epigenetic changes in innate immune cells, some also express virulence factors that can directly or indirectly alter chromatin modifications in order to persist within the host. Chromatin modifications have different abilities to be preserved over time and transmitted through cell division. For example, some histone modifications such as acetylation and phosphorylation are strongly induced by stimulation with extracellular signals but are short-lived. Histone methylation however is more stable, lasting for hours or days. Finally, DNA methylation is the most durable modification identified and can be passed on to daughter cells (Pereira, Hamon, & Cossart, 2016).

Some bacteria have been shown to trigger long-lasting epigenetic modifications. Examples include Mycobacterium leprae that can reprogram Schwann cells (fully differentiated cells of the peripheral nervous system) into progenitor/stem-like cells (Masaki et al., 2013). Helicobacter pylori-induced chronic inflammation, altering the expression of Il1b, nitric oxide synthase 2 (Nos2) and Tnf, is associated with the generation of aberrant DNA methylation at the promoter of genes methylated in gastric cancer and has, therefore, been linked with carcinogenesis (Maeda, Moro, & Ushijima, 2017). Numerous other bacteria have been reported to induce histone modifications that play a role in regulation of host transcription. These mechanisms typically rely on bacterial and host epigenetic factors such as methyltransferases and histone deacetylases (HDAC). The lasting impact of such effects remains understudied. Examples of these bacteria-mediated host epigenetic modifications are summarized below and in more detail in Table 1.

Table 1:

Bacterial epigenetic modulations and their effects on the immune response

Bacterium	Training/activating effects on immune response		Repressive effects on immune response
	Factor	Effects/Mechanism	Factor	Effects/Mechanism
Anaplasma phagocytophilum	N/A	N/A	Live bacterium	HDAC1 activation decreased H3 acetylation downregulation of host defense genes (Garcia-Garcia, Barat, Trembley, & Dumler, 2009)
Bacillus anthracis	N/A	N/A	Bacillus anthracis suppressor of variegation, enhancer of zeste, trithorax (BaSET)	methylation of histone H1 repression of NF-κB functions decreased IL-6, IL-8 and TNF-α production increased virulence and bacterial burden compared to mutant lacking BaSET (Mujtaba et al., 2013)
			Lethal toxin (LT)	impaired H3S10 phosphorylation impaired IL-6 and KC production decreased IL-8 production (Raymond et al., 2009)
Bacillus Calmette-Guerin (live attenuated Mycobacterium bovis)	Live attenuated bacterium	increased IL-1β production in human monocytes decreased viremia upon secondary challenge with yellow fever virus (R. J. Arts et al., 2018) increased glycolysis, glutamine metabolism and oxidative phosphorylation in monocytes (R. J. Arts et al., 2016) rewiring of HSPCs transcriptional programs towards myelopoiesis HNF1a and HNF1b regulation (Cirovic et al., 2020)	N/A	N/A
Chlamydia trachomatis	N/A	N/A	Nuclear effector (NUE)	methylation of histones H2B, H3 and H4 (Pennini et al., 2010)
Legionella pneumophila	N/A	N/A	Regulator of methylation A (RomA)	H3K14 trimethylation decreased Tlr5 and Il6 expression increased intracellular bacterial replication (Rolando et al., 2013)
Listeria monocytogenes	Protein and/or nucleic acid of heat killed bacterium	increased Mpge1, Tnfα and Nos2 expression decreased sod2 expression decreased OxPhos decreased bacterial burden (Luukinen et al., 2018)	Listeriolysin O (LLO)	dephosphorylation of H3S10 decreased CXCL2 expression (Hamon et al., 2007)
			Internalin B (InlB)	translocation of SIRT2 to the nucleus H3K18 deacetylation increased infection rates (Bierne & Hamon, 2020)
			Listeria nuclear targeted protein A (LntA)	target BAHD1 induces type III IFN gene expression increased infection rates (Lebreton et al., 2011)
Mycobacterium tuberculosis	N/A	N/A	Rv1988	methylation of H3R42 repression of genes involved in ROS production increased bacterial survival (Yaseen et al., 2015)
Neisseria gonorrhoeae	N/A	N/A	Histone deacetylase-like enzyme (Gc-HDAC)	reduced expression of host defense peptides reduced H3K9ac marks at promoters of pro-inflammatory genes (Zughaier et al., 2020)
Pseudomonas aeruginosa	Purified flagellin	increased IL-8 and IL-6 secretion upon secondary challenge with Aspergillus fumigatus (Bigot et al., 2020)	2-aminoacetophenone (2-AA)	increased HDAC1 activity deacetylation of H3K18 at TNF-α promoter decreased pro-inflammatory cytokines increased anti-inflammatory cytokines chronic infection (Bandyopadhaya et al., 2016)
			Purified flagellin	decreased IL-8 and IL-6 secretion upon secondary challenge with Pseudomonas aeruginosa, Stenotrophomonas maltophilia and LPS (Bigot et al., 2020)
Shigella flexneri	N/A	N/A	OspF	dephosphorylation of mitogen-activated protein kinases impaired H3S10 phosphorylation decreased production of IL-8 reduced leukocyte recruitment (Arbibe et al., 2007)
Staphylococcus aureus	Live bacterium	decreased dermonecrosis decreased bacterial burden (Chan et al., 2017; Wong Fok Lung et al., 2020)	N/A	N/A
Streptococcus pneumoniae	Live bacterium	altered phenotype and creatine metabolism in alveolar macrophages decreased Ccl22, Retnla, Marco, Tnf expression in alveolar macrophages decreased bacterial burden (Guillon et al., 2020)	Pneumolysin (Ply) and Pyruvate oxidase (SpxB)	dephosphorylation of protein phosphatase 1 dephosphorylation of H3S10 increased intracellular bacterial infection (Dong et al., 2020)

N/A: not available; HDAC1: histone deacetylase 1.

Numerous bacterial proteins have the ability to induce host epigenetic modifications. These include the lethal toxin of Bacillus anthracis, perfringolysin O from Clostridium perfringens, aerolysin from Aeromonas hydrophila, pneumolysin of Streptococcus pneumoniae and listeriolysin O from Listeria monocytogenes that mediate histone dephosphorylation and deacetylation through pore formation, resulting in impaired cytokine production and immune cell recruitment (Bierne & Hamon, 2020; Hamon et al., 2007; Raymond et al., 2009). Some bacterial factors also act by directly targeting host factors involved in epigenetic remodeling. For example, Anaplasma phagocytophilum and Pseudomonas aeruginosa have been shown to increased HDAC1 expression, resulting in decreased histone acetylation, decreased cytokine production and increased bacterial burdens (Bandyopadhaya, Tsurumi, Maura, Jeffrey, & Rahme, 2016; Cabezas-Cruz et al., 2016). One class of proteins identified across several bacterial species are the suppressor-of-variegation, enhancer-of-zeste, trithorax (SET) methyltransferase proteins. These proteins act by methylating histones and preventing their acetylation and have been identified in Chlamydia trachomatis (Pennini, Perrinet, Dautry-Varsat, & Subtil, 2010), Legionella pneumophila (Hawn et al., 2003; Rolando et al., 2013) and Bacillus anthracis (Mujtaba et al., 2013). Mycobacterium tuberculosis also expresses a methyltransferase, Rv1988, which targets histone H3 repressing reactive oxygen species production, thereby giving a survival advantage to the bacterium (Yaseen, Kaur, Nandicoori, & Khosla, 2015). Finally, some pathogenic bacteria encode their own histone-modifying enzymes, the human pathogen Neisseria gonorrhoeae, encodes a HDAC-like enzyme: Gc-HDAC. This factor was found to reduce the expression of host defense peptides cathelicidin LL-37, human beta-defensin 1 (HBD-1) and secretory leukocyte protease inhibitor (SLPI) in human macrophages via epigenetic modifications of host innate immune genes (Zughaier, Rouquette-Loughlin, & Shafer, 2020).

While all these studies demonstrate the ability of pathogenic bacteria to directly and indirectly affect the host epigenome, little is known about the long-term impacts of such events on innate immune cells. It is possible, given their repressive effects on pro-inflammatory gene expression and epigenetic alterations that these events would result in immune paralysis and tolerance to subsequent challenges with related or unrelated pathogens. Many of these studies have been performed in vitro, so it is also unknown if these would have an impact in vivo and ultimately interfere with a training phenotype. It is also plausible that the observed effects are only transient and would dissipate once the infection is resolved. Studying these virulence factors from an innate immune memory perspective has the potential to yield informative answers and in many cases might open up new potential therapeutic approaches.

Perspectives

Immunotherapies, or the use of substances that stimulate the immune system to prevent or cure diseases, have been in place for many years. The use of innate immune memory to combat infections is a more recent concept with promising potential. As an example, trained immunity-based treatments could be administered to prevent the spread of SARS-Co-V2, the agent of the current COVID-19 pandemic until a vaccine is designed. Several studies have suggested that countries where BCG vaccination is given at birth have a lower infection rate and fewer COVID-19-related deaths (Covián, Retamal-Díaz, Bueno, & Kalergis, 2020; Escobar, Molina-Cruz, & Barillas-Mury, 2020; O’Neill & Netea, 2020) and that enhancement of innate immune responses by BCG is not associated with increased disease or severity of symptoms during the COVID-19 pandemic (Aksu, Naziroğlu, & Özkan, 2020; Moorlag et al., 2020). While these results indicate a protective effect of BCG vaccination, definite proof of causality cannot be obtained due to differences in demographic and genetic structure of the populations, differences in non-pharmaceutical preventive measures (such as quarantine and social distancing) and differences in diagnosis and reporting of COVID-19 cases. Therefore, randomized controlled trials are required to support this hypothesis. Clinical trials using BCG vaccination as a means to prevent or reduce the severity of SARS-CoV-2 infection are underway in several countries including the Netherlands, Australia, Denmark, the United States, France and Uruguay (Netea, Giamarellos-Bourboulis, et al., 2020).

The use of trained immunity has also been proposed for the development of personalized cancer immunotherapies: a mechanism comparable to LPS-mediated immune paralysis has been observed in cancer where the development of myeloid-derived suppressor cells (MDSCs) is observed. This heterogeneous progenitor cell population mediates immunosuppressive effects within the tumor microenvironment that attenuate the anti-tumor response and promote tumor growth and metastasis (Hou, Hou, Huang, Lei, & Chen, 2020). The functions of MDSCs are dependent on specific epigenetic modifications including DNA methylation and histone modifications (C. Zhang, Wang, Liu, & Yang, 2016). Therefore, new immunotherapies using trained immunity could be designed to induce the epigenetic rewiring of these cells (Mulder, Ochando, Joosten, Fayad, & Netea, 2019).

Another potentially appealing aspect of trained immunity is its effect on HSPCs. Indeed, increased myelopoiesis could be a desirable feature for neutropenic patients or to counteract the effects of chemotherapy-induced myelosuppression (Mitroulis et al., 2018). Specific in vivo training of progenitor cells could also be a way to induce long-lasting protective responses against related and unrelated pathogens. Although trained immunity has been shown to last for up to one year (Netea, Domínguez-Andrés, et al., 2020), the extent of this persistence needs to be further characterized and what dictates the duration. Studies so far have examined a single infection, or single dose of immune stimulant and its effects on innate training. Whether an additional dose several months later would reinforce these epigenetic changes or induce a tolerized phenotype are yet to be explored. Since the long-lasting effects of this phenomenon seem to be carried out by epigenetic modifications, a potential way to prolong its effects could be the inhibition of host factors that remove these epigenetic marks, such as HDACs.

Trained immunity can also have downsides, by triggering hyperinflammation and contributing to various disorders including atherosclerosis, diabetes, chronic inflammation and neurodegenerative disorders (Włodarczyk, Druszczyńska, & Fol, 2019). Several options have been proposed control the undesirable effects of trained immunity. Modulation of pathogen- and danger-associated molecular patterns (PAMPs and DAMPs) pathways by molecular inhibitors can be used to regulate this phenomenon. Nanocarriers have been suggested as a way to deliver these molecules specifically to their target cells (Mulder et al., 2019). Targeted inhibition of metabolic pathways involved in trained immunity, such as glycolysis or glutaminolysis can also be used to counteract its deleterious effects (R. J. Arts et al., 2016; Cheng et al., 2014). The use of histone and DNA methylation inhibitors to suppress epigenetic modifications could also prevent such activation. As summarized in this review, numerous bacteria produce factors that directly inhibit enzymes involved in host histone and DNA remodeling therefore, some of these factors could potentially be used to counteract the effects of trained immunity.

The bacterial virulence factors that have a deleterious impact on the host epigenome are well characterized and a better definition of the bacterial triggers inducing trained immunity will allow the development of products that better regulate this phenomenon. In the context of BCG immunization, NOD2 has been identified as the receptor by which trained immunity is mediated (Kleinnijenhuis et al., 2012), β-glucan inoculation triggers epigenetic remodeling of macrophages via dectin 1-dependent pathways (Cheng et al., 2014) and flagellin has also been identified as a potential trigger of trained immunity (Bigot et al., 2020) but the exact mechanisms by which these actions are performed still remain to be fully elucidated. Once, these components and mechanisms have been further characterized, trained immunity-based immunotherapies could have tremendous applications. Given the unspecific nature of trained immunity-based responses, immunization with a single trained-immunity-based vaccine could protect against a range of infections, compared to the antigen-specific nature of traditional vaccines [reviewed here (Sánchez-Ramón et al., 2018)]. While not exhaustively investigated, some innate stimuli appear to confer heterologous protection in the context of trained immunity. The use of bacterial strains lacking those factors influencing host epigenetics might prove useful in an innate immune training. Alternatively, identifying the bacterial components or PAMPS that induce training could be used as a protein or peptide-based vaccine. Trained immunity stimuli could also be incorporated as adjuvants of common vaccines in order to boost both innate and adaptive memory responses simultaneously. The use of non-infectious products to initiate trained immunity responses could be of particular interest in the context of at risk populations to prophylactically increase immune defenses against specific pathogens or to prevent/resolve immune paralysis in the context of infectious and non-infectious conditions (such as sepsis or trauma).