The role of neutrophils in trained immunity

The principle of trained immunity represents innate immune memory due to sustained, mainly epigenetic, changes triggered by endogenous or exogenous stimuli in bone marrow (BM) progenitors (central trained immunity) and their innate immune cell progeny, thereby triggering elevated responsiveness against secondary stimuli. BM progenitors can respond to microbial and sterile signals, thereby possibly acquiring trained immunity‐mediated long‐lasting alterations that may shape the fate and function of their progeny, for example, neutrophils. Neutrophils, the most abundant innate immune cell population, are produced in the BM from committed progenitor cells in a process designated granulopoiesis. Neutrophils are the first responders against infectious or inflammatory challenges and have versatile functions in immunity. Together with other innate immune cells, neutrophils are effectors of peripheral trained immunity. However, given the short lifetime of neutrophils, their ability to acquire immunological memory may lie in the central training of their BM progenitors resulting in generation of reprogrammed, that is, “trained”, neutrophils. Although trained immunity may have beneficial effects in infection or cancer, it may also mediate detrimental outcomes in chronic inflammation. Here, we review the emerging research area of trained immunity with a particular emphasis on the role of neutrophils and granulopoiesis.

develop long-term specific immunological memory. 2 However, this traditional dichotomy may not be able to explain the full complexity of immune responses.
After the discovery of pattern recognition receptors (PRRs) in innate immunity evidence accumulated for the existence of protection mechanisms against infection and potentially reinfection that could not be solely attributed to the adaptive branch of immunity. 3,4 PRRs, such as Toll-like receptors (TLRs), NOD-like receptors (NLRs), RigIhelicases, and C-type lectin receptors, are expressed in a plethora of innate immune cells, allowing them to recognize and respond to multiple pathogen-expressed molecules, that is, pathogen-associated molecular patterns (PAMPs), and damage-associated molecular patterns (DAMPs). [5][6][7] Importantly, studies in plants and invertebrates that do not have adaptive immunity, as well as significant recent work in vertebrates, and epidemiological studies in humans suggest that innate immune cells may also experience long-lasting alterations that may confer increased protection against reinfection. This property has been designated trained innate immunity (or trained immunity) and is fundamentally based on innate immune memory. [8][9][10][11] Unlike the classical immunological memory of adaptive immunity that requires gene rearrangement and clonal expansion of antigenspecific lymphocytes, the enhanced innate immune responses to future homologous or even heterologous challenges seen in trained immunity depend on sustained epigenetic modulation of innate immune cells that modify their transcriptional program and function and increase their immune preparedness. 9,10,12 Therefore, innate immune memory refers to an altered functional state of the cell that persists from weeks to months; however, there are studies showing also heterologous protection induced by live vaccines even after longer periods, for example, five years. 9,13 In other words, trained innate immune cells display an acquired capacity to react qualitatively different, mostly stronger, upon being challenged by secondary stimuli, as compared to untrained cells. This altered responsiveness is attributed to the differential regulation of the expression of several inflammatory genes. The mechanisms governing this differential gene expression are only partially understood and may involve alterations in chromatin remodeling, histone modifications, and rewiring of cellular metabolism, 9 as will be outlined below.

| INDUCER S OF TR AINED IMMUNIT Y
Initially, it was thought that trained immunity is induced mainly by exogenous microbial stimuli. Currently, it is appreciated that nonmicrobial and even endogenous stimuli may induce trained immunity. 9 Microbial stimuli that are well known to mediate innate immune training are live attenuated vaccines, such as the Bacille Calmette-Guerin (BCG) vaccine, 14 the oral polio vaccine, 15 the measles vaccine, 16 the smallpox vaccine, 17 and the new vaccine against tuberculosis MTBVAC. 18 Additionally, prototype and most studied trained immunity agonists are β-glucans, 9 which are highly conserved glucose polymers from the wall of various fungi and yeast and act as immunostimulatory agents. 19 Moreover, lipopolysaccharide (LPS) has also been implicated in mediating trained immunity. 20,21 Hepatitis B virus 22 and the pathogen Plasmodium falciparum 23 have also been described to induce innate immune training.
In addition to the aforementioned microbial stimuli and microbiota-derived immunostimulatory factors, 24 trained immunity effects have been also described upon exposure to endogenous nonmicrobial factors such as lipoproteins, 25,26 uric acid, 27,28 aldosterone, 29 catecholamines, 30 S100-alarmins, 31 heme, 32 activators of the liver X receptor pathway, 33 or interferons. 34 Essentially, the list of trained immunity agonists is likely very wide, since several factors that act as PAMPs or DAMPs might bear the potential to induce innate immune training. 11 However, the inflammatory effector function of each stimulus and the underlying pathways involved may vary substantially, thereby differentially affecting the capacity of distinct stimuli to induce trained immunity-associated long-term changes in cells. Hence, trained immunity-related programs may display substantial versatility and complexity.

| EPIGENETIC AND IMMUNOMETABOLIC REG UL ATI ON OF TR AINED IMMUNIT Y
Pro-inflammatory gene loci in quiescent innate immune cells are mostly found to be repressed. 35 Several findings suggest that the memory-like phenotypes in innate immunity are associated with altered chromatin accessibility in trained immune cells or their BM progenitors, as assessed by assay for Transposase-Accessible Chromatin with high-throughput sequencing (ATAC-seq) 36,37 and may also involve sustained stimulus-induced histone modifications at the level of inflammatory genes, thereby altering their transcription and the responsiveness of the cells upon future challenges. 9,12 Some important epigenetic marks at distal regulatory elements or at the promoter of genes have been related to trained immunity programs: the histone 3 lysine 27 acetylation (H3K27ac), the histone 3 lysine 4 monomethylation (H3K4me1), and the histone 3 lysine 4 trimethylation (H3K4me3). The H3K27ac is found at distal enhancers that carry a H3K4me1 and the H3K4me3 marks promoters of activated genes. 9 The collision of transcription factors on these regions is described to be random. 38 However, this stochastic regulation of gene expression is not optimal in cases that require immediate responses to stimuli. Studies performing analysis of chromatin interactions have reduced this notion of stochasticity in gene transcription by showing that multiple rapidly responding genes in innate immune cells are organized in topologically associated domain (TAD) structures and interact within these complexes. 9,39 These structures may enable immune-related genes to interact with some long noncoding RNAs (lncRNAs), the so-called "immune-gene priming lncRNAs" (IPLs) and prime their transcriptional activation. 40 IPLs are located in the same TAD as the immune genes they regulate, thereby directing histone modulating enzyme complexes to the promoters of this set of co-regulated genes, and allowing for their epigenetic reprogramming. 41 The prototypic IPL, UMLILO (Upstream Master LncRNA of the Inflammatory chemokine Locus) forms chromosomal contacts with the ELR+ CXCL chemokines (Interleukin [IL]-8, CXCL1, CXCL2, and CXCL3) and acts in order to facilitate their H3K4me3 epigenetic priming. 40 It is also suggested that other cytokines, such as IL-6 and IL-1β, that are involved in trained immunity, are similarly regulated in an IPL-dependent manner. 42 In addition, βglucan-mediated innate immune training induces increased nuclear factor of activated T cells (NFAT)-dependent transcription of IPLs that subsequently results in innate immune gene epigenetic rewiring. 42 Furthermore, priming with β-glucan or BCG can upregulate IPLs in BM progenitor cells that in turn results in greater deposition of H3K4me3 marks on immune gene promoters. 42 Hence, lncRNA contributes to induction of trained immunity and may serve as an entirely new druggable target for therapeutic immunomodulation.
Epigenome modification is intimately linked with cellular metabolism, as several metabolic pathways produce metabolites that may modify chromatin and the activity of epigenetic factors. 43,44 Given the relevance of epigenetic reprograming in trained immunity, it is not surprising that changes in cell metabolism may also mediate the establishment of innate immune memory. 45 Changes in cellular metabolic pathways, for example, glycolysis, cholesterol biosynthesis, amino acid and fatty acid metabolism and oxidative phosphorylation, can dictate innate immune cell plasticity, regulate the activity of chromatin-modifying enzymes, and have been implicated in trained innate immunity. [45][46][47] For example, an increase in mechanistic target of rapamycin (mTOR)-and hypoxia-inducible factor-1α (HIF1α)-mediated aerobic glycolysis is found in both β-glucan and BCG trained monocytes. 48 In addition, BCG-induced trained immunity requires a strong increase in glycolysis that is responsible for histone modifications and functional changes in monocytes. 49 Induction of trained immunity in human monocytes is also linked to elevated glutaminolysis and accumulation of the tricarboxylic acid (TCA) cycle metabolite, fumarate, that regulates epigenetic changes by inhibiting KDM5 histone demethylases. 50 Studies in human monocytes revealed that β-glucan-dependent training results in increased mevalonate levels, thereby promoting trained immunity through activation of the insulin-like growth factor 1 receptor and mTOR pathway. 51 Furthermore, another TCA cycle metabolite, alpha-ketoglutarate promotes an anti-inflammatory M2-like polarization of macrophages; reducing the ratio of alpha-ketoglutarate to succinate in turn supports a pro-inflammatory M1-like phenotype in macrophages. 52 The metabolite itaconate that is produced from a TCA metabolite, cis-aconitate, and has been ascribed multiple antiinflammatory actions in mouse and human macrophages, [53][54][55] has been implicated as a target of β-glucan-induced trained immunity in monocytes. Specifically, β-glucan-induced trained immunity downregulates expression of the enzyme that regulates itaconate production, immune-responsive gene 1 (IRG1) in monocytes, thereby also counteracting development of immune tolerance. 56 Together, there is a close link between immunometabolic and epigenetic pathways during establishment of trained immunity; the better understanding of this link and its regulation requires substantial future investigation. Of note, research on the immunometabolic regulation of trained immunity has mostly focused on monocytes and macrophages, whereas neutrophils have been poorly studied in this regard so far.

| BONE MARROW HEMATOP OIE S IS AND CENTR AL TR AINED IMMUNIT Y
Initially, trained immunity was studied in mature myeloid cells.
However, the short lifetime of these cells, especially monocytes and neutrophils, in the circulation was in disagreement with the longterm effects of innate immune memory persisting for months to years. 9 This discrepancy could be reconciled by findings that trained immunity does not only occur in mature myeloid cells within peripheral tissues and organs (peripheral trained immunity) but may in fact be initiated in progenitors of these cells in the BM (central trained immunity) 9,57 ( Figure 1).
All mature blood cells, including innate immune cells, are generated in the BM by a hierarchically organized process called hematopoiesis. 58 The hematopoietic stem cells (HSCs) are on the top of the hematopoiesis pyramid and display the unique capacity to self-renew. Moreover, HSCs have the ability to differentiate to all blood lineages. When HSCs are activated, they differentiate into multipotent progenitors (MPPs). HSCs and MPPs together form the pool of hematopoietic stem and progenitor cells (HSPCs) and are the main two populations within the lineage − Sca1 + c-kit + (LSK) cells in the BM. 59,60 For the generation of myeloid cells, HSPCs will subsequently differentiate into more committed progenitors, such as common myeloid progenitors (CMPs) and granulocyte-macrophage progenitors (GMPs). 61 Under infectious or inflammatory conditions with increased demand for myeloid cells in the periphery, myeloid progenitors in the BM expand; in addition, HSCs also respond with expansion and enhanced myeloid differentiation; this process giving rise to new myeloid cells upon stress conditions is designated emergency myelopoiesis. 57,60,61 When the aforementioned process leads to predominantly increased numbers of neutrophils (e.g., in the context of an acute bacterial infection), it is specifically termed emergency granulopoiesis. 61 In this regard, it is important that HSCs and HSPCs are capable to recognize and respond to a variety of soluble factors reaching the BM microenvironment during peripheral infection or inflammation. 57,62 HSPCs express PRRs, such as TLRs that recognize PAMPs, as well as receptors for pro-inflammatory cytokines, such as IL-1, IL-6, tumor necrosis factor (TNF), type I and II interferons (IFN) and receptors for growth factors, such as granulocyte macrophage-colony stimulating factor (GM-CSF) or macrophagecolony stimulating factor (M-CSF), enabling them to recognize systemic inflammatory challenges and acutely respond. 57,[63][64][65] The HSPC response to such PAMPs, cytokines, or growth factors commonly results in emergency myelopoiesis with an increase in the GMP pool and subsequent myeloid cell production. 61,66 Importantly, however, although HSPC reactions are critical for the response to infections as well as for restoration of hematopoiesis after irradiation, excessive and chronic activation may lead to impaired function, or even exhaustion, of HSPCs resulting in chronic pathologies. [67][68][69] In the context of trained innate immunity and its early stages, our group and others have provided evidence that different agonists of trained immunity can induce responses in HSPCs that resemble emergency myelopoiesis processes. Via engaging signaling triggered by different pro-inflammatory factors, such as IL-1, IFNs, and growth factors, such as GM-CSF, agonists of trained immunity, such as BCG or β-glucan, may induce long-lasting alterations in HSPCs, at the metabolic and epigenetic level, which lead to sustained increase of myelopoiesis and granulopoiesis. 20,36,37,[70][71][72][73][74][75] In addition, this central trained immunity leads to generation of trained myeloid cells, that is, cells with increased inflammatory responsiveness (Figure 1).
The existence of central BM-mediated trained immunity as a major arm of innate immune memory was established by performing BM transplantation studies in mice. 37,71 Increased myelopoiesis and granulopoiesis as well as generation of myeloid cells with enhanced inflammatory potential are both integral components of central BMmediated trained immunity that results in sustained elevation of systemic innate immune responses upon future challenges. 20,36,37,[70][71][72][73][74][75]

| B ENEFICIAL VER SUS MAL ADAP TIVE INNATE IMMUNE TR AINING
A large body of published work demonstrates the beneficial role of trained immunity in mice and humans especially during the response to infectious pathogens. 9 Epidemiological findings report that several vaccines are able to promote "non-specific" immunity by protecting the immunized individual not only against the target pathogen but also against a broad variety of unrelated infections. 76 As an example of the off-target beneficial effects of vaccines, BCG vaccination in children of West Africa resulted in overall reduced mortality, due to protection against other infections in addition to tuberculosis. 77 Other live vaccines, such as the oral polio vaccine, the measles vaccine, and the smallpox vaccine, have been also shown to induce heterologous protective effects. 76 Trials performed in children 78 and adults 14,79 demonstrated an elevated activation of innate immune cells following BCG vaccination. Accordingly, BCG vaccination in humans resulted in increased pro-inflammatory function of monocytes that was accompanied by partial protection against experimental malaria 80 and yellow fever. 81 In mice, BCG vaccination provides protection against secondary infection with Schistosoma mansoni, 82 Candida albicans, 83 and Mycobacterium tuberculosis. 75 BCG administration protected SCID mice that do not have adaptive immune responses from lethal candidiasis, 14 thereby further underlining the functional relevance of innate immune memory in this process.
Another commonly used trigger of trained immunity, β-glucan, also confers protection against future infectious challenges.
Specifically, mild infection with Candida albicans or simply pretreating with β-glucan could protect against subsequent lethal candidiasis; this phenomenon relied on induction of trained immunity and was independent of adaptive immunity, as shown by the use of Rag1-deficient mice lacking lymphocytes. 84 Moreover, Candida albicans gut colonization induced trained immunity that led to enhanced immune responses against multiple, systemic infections. 85 F I G U R E 1 Central and peripheral trained immunity. Central trained immunity defines that the process of myelopoiesis is trained by epigenetic rewiring of HSPCs (HSCs and MPPs) as well as of GMPs, thereby giving rise to more mature myeloid cells (neutrophils and monocytes) and importantly to myeloid cells with higher inflammatory potential. Agonists, such as the β-glucan or the BCG vaccine, may promote central trained immunity in a manner that involves signaling by factors, such as GM-CSF, IL-1β, type I, or type II IFNs. Induction of trained immunity also involves the direct modulation of the inflammatory preparedness of mature myeloid cells in the bloodstream or peripheral organs (peripheral trained immunity). BCG, Baccillus Calmette-Guerin; GMP, granulocyte-macrophage progenitor; HSC, hematopoietic stem cell; MPP, multipotent progenitor.
Additionally, β-glucan-induced training has been reported to exhibit a protective effect against systemic Streptococcus pneumoniae infection 86 and Mycobacterium tuberculosis infections. 74 In addition, there is evidence supporting a possible anti-tumor effect of trained immunity agonists BCG 87 and β-glucan. 88,89 BCG has been reported as a potential therapeutic strategy for several malignancies, such as bladder cancer, where it has been an approved therapy against non-muscle-invasive bladder cancer forms, 90 melanoma, 91 leukemia, 92 and lymphoma. 93 With regard to β-glucan, it is currently being used in a series of clinical trials as an adjuvant immunotherapy against different forms of cancer. 89 In this regard, and as it will be outlined in detail below, we could recently show that the anti-tumor effects of β-glucan are, at least partially, mediated by the induction of central trained immunity, specifically trained granulopoiesis, that led to generation of neutrophils with potent anti-tumor activity. 37 Taken together, multiple studies support that induction of trained immunity may have beneficial effects in the context of heterologous infections or may activate anti-cancer programs.
While innate immune memory involves changes that improve the hosts' protection against infection or cancer, trained innate immunity may have also a dark side. The elevated inflammatory responsiveness associated with trained immunity-induced innate immune cell rewiring may also result in maladaptive or detrimental effects during chronic inflammation and tissue damage. A cardinal example is sterile inflammation caused by Western diet that may induce trained immunity and chronic cardiometabolic diseases. 57,94 In addition, innate immune memory-induced augmented immune cell activation could provide mechanistic underpinnings for the epidemiological observations linking infections with atherosclerosis occurrence. 95,96 Several endogenous stimuli have been demonstrated to trigger trained immunity, such as oxidized low-density lipoprotein (oxLDL), thereby eliciting increased inflammatory cytokine production, which contributes to the pathogenesis of atherosclerosis. 25,26 Moreover, Western diet-induced NLRP3 inflammasome-dependent epigenetic and transcriptomic reprogramming of GMPs results in augmented inflammatory responses even after switching mice back to normal diet. 36 Accordingly, monocytes obtained from individuals with hypercholesterolemia exhibit enhanced inflammatory properties consistent with trained immunity, which persisted even after statin treatment. 97 Furthermore, monocytes in the circulation of patients with symptomatic atherosclerosis exhibit increased production of cytokines upon future challenge, accompanied by metabolic and epigenetic rewiring. 98 Interestingly, we could recently show that maladaptive BMmediated trained immunity may provide a mechanistic link for the development of inflammatory comorbidities. Specifically, experimental periodontitis triggered IL-1-dependent maladaptive central trained immunity, associated with epigenetic rewiring of HSPC and sustained generation of monocytes and neutrophils with enhanced inflammatory responsiveness that could thereby exacerbate inflammatory arthritis, 72 which is a frequent periodontitis comorbidity. 99 Importantly, periodontitis-driven maladaptive myelopoiesis training was transmissible by BM transfer to recipient mice, which also exhibited increased susceptibility to inflammatory arthritis. 72 Since trained immunity lacks specificity to the initial stimulus, maladaptive training of myelopoiesis may underlie the emergence not only of the periodontitis-arthritis comorbidity but also of several inflammatory comorbidities. 60,99,100 Collectively, trained innate immunity is a double-edged sword that can promote both beneficial and deleterious effects in a contextdependent manner. 60,72,[99][100][101] 6 | NEUTROPHIL S AND G R ANULOP OIE S IS Neutrophils, also known as polymorphonuclear (PMN) leukocytes, comprise the majority of circulating leukocytes in humans.
Neutrophils have an enormous turnover ratio with roughly 0.5-1 x 10 11 neutrophils being daily generated under steady-state conditions. 102 The de novo production of neutrophils occurs in the BM by myeloid progenitors, GMPs through a process termed as granulopoiesis. 61,[103][104][105][106] As neutrophils are the first cells to respond in infection or injury, the demand for these cells in the periphery is constantly high. Thus, adaptation of myeloid progenitors in such cases underlines the process of emergency granulopoiesis, which leads to production of approximately 10 12 neutrophils daily. 61 Given these large amounts of neutrophils, it is not surprising that approximately 60% of the BM hematopoiesis is dedicated to producing these cells. 107 Recent studies revealed further details regarding the differentiation paths from GMPs to neutrophils.
Three subpopulations were found in the BM, neutrophil precursors that differentiate into immature neutrophils and then mature neutrophils. The differentiation of GMPs to neutrophil precursors was dependent on C/EBPε transcription factor and involves progenitors within the GMP population. 105 Novel research has recently shed more light into the different subpopulations of neutrophil precursors found in humans, such as proNeus, 104 preNeus, 105 early neutrophil progenitors, 108 human neutrophil progenitors, 109 and neutrophil-committed progenitors. 110 Interestingly, such committed progenitors and neutrophil precursors expand in infection and cancer. 104,105 Although neutrophils comprise a terminally differentiated population, 111 single-cell analytical studies have revealed substantial heterogeneity within neutrophils. 106 The general notion is that these cells are short lived; however, their lifespan may vary in different tissues. Initially, ex vivo survival assays and experiments utilizing adoptive transfer demonstrated that neutrophils live 8-12 h in the bloodstream and 1-2 days in the tissue, 102,112,113 whereas in vivo methods indicate that the neutrophil lifetime in humans may reach 5 days. 114 At the end of their lifetime, neutrophils undergo apoptosis and they subsequently get phagocytically cleared by macrophages through a process named efferocytosis. 115 Aged neutrophils may return to the BM and get cleared by BM macrophages. 116,117 Despite their impressive turnover rate, the number of neutrophils in the circulation remains constant. This is owing to the well-orchestrated balance between production and elimination. 118 Their multiple phenotypic changes and properties suggest existence of great neutrophil functional heterogeneity. 119 The functional heterogeneity of neutrophils in the periphery may be shaped by the tissue microenvironments. For instance, neutrophils obtain in peripheral tissues, for example, in the lung, the capacity to contribute to vessel repair or to promote the integrity of hematopoiesis, hence facilitating the restoration of hematopoiesis upon irradiation. 103 Moreover, neutrophils display a great repertoire of functions, such as degranulation, phagocytosis, antigen-presentation, formation of neutrophil extracellular traps (NETs), release of cytokines and other inflammatory and anti-microbial factors, 119,120 thereby actively regulating inflammation and its resolution, 121 and participating in several pathologies, including cancer. 122,123 Moreover, as it was recently demonstrated and will be outlined in the next paragraphs, neutrophils are integral components of innate immune memory ( Figure 2). 37,72,124 7 | NEUTROPHIL S AND G R ANULOP OIE S IS IN TR AINED INNATE IMMUNIT Y As alluded to above, neutrophils have been recently recognized as effectors of trained immunity. 37,124 The BCG vaccine, a prototypical agonist of trained immunity, provides broad protection against infection by multiple unrelated pathogens. 9 Although these off-target protective effects of BCG are typically attributed to innate immune training of monocytes, recent studies suggest a prominent role of neutrophils in this regard. 124,125 A few weeks following BCG vaccination of healthy humans, the absolute number of neutrophils in the blood increases. 124,125 Nevertheless, 3 months after vaccination, the neutrophil counts re-equilibrate back to normal; however, neutrophils maintain qualitative changes and particularly an activated/ trained phenotype with increased expression of CD11b, CD66b and diminished expression of CD62L and PDL1. 124 This highly activated phenotype is accompanied by enhanced capacity of BCG-trained F I G U R E 2 Neutrophils in protective and maladaptive trained immunity. Trained innate immunity and particularly trained granulopoiesis may exert both beneficial and detrimental effects. Trained immunity leads to sustained reprogramming of granulopoiesis with generation of trained neutrophils that display more potent effector mechanisms such as degranulation, ROS production, phagocytosis, antigen presentation, release of cytokines and other immune mediators, and increased expression of adhesion molecules and activation markers. Moreover, the microbiota may promote an aged phenotype in mouse neutrophils that is characterized by decreased expression of CD62L, increased expression of CXCR4, 136 and decreased proliferation. 140 Compared to young neutrophils, aged neutrophils display increased expression of factors involved in adhesion and migration, such as lymphocyte function-associated antigen-1 (LFA-1), macrophage-1 antigen (Mac-1), PECAM-1/CD31, CD44, very-lateactivation antigen-4 (VLA-4), ICAM-1, as well TLR-4. 140 Consistent to a more active phenotype, microbiota-induced aged neutrophils stimulated with LPS produce increased levels of ROS and NETs. 136 The hyperactive phenotype of aged neutrophils and the increased NET production contribute to exacerbated organ damage of mice in the context of endotoxin-induced inflammation. 136 Additionally, aged neutrophils promote the pathology of sickle cell disease in an animal model; accordingly, the depletion of microbiota limits the expansion of aged neutrophils and improves disease outcome. 136 In another study, mouse-aged neutrophils, exposed to PAMPs (LPS or lipoteichoic acid) or DAMPs (e.g., high-mobility group protein 1; HMGB1), upregulate Mac-1 and TLR-4 and present a significantly higher phagocytic potential, as compared with non-aged neutrophils. 140 Recently, it was demonstrated that COVID-19 infection can also induce central trained immunity with increased granulopoiesis. As outlined above, innate immune training may also be induced by nonmicrobial DAMPs besides PAMPs. 9 For instance, heme, which acts as a DAMP, may trigger trained immunity. Administration of heme to mice promotes a myeloid bias by increasing the frequency of myeloid-biased multipotent progenitors (MPP3) and decreasing the lymphoid-biased multipotent progenitors (MPP4), associated with higher neutrophil numbers in the BM. 32 Furthermore, heme induces long-lasting epigenetic changes in the LSK compartment. 32 Heme-induced trained immunity results in higher neutrophil and monocyte infiltration into the peritoneum following a secondary LPS challenge. 32 Trained immunity induced by heme is protective against polymicrobial sepsis when heme is given 7 but not 28 days prior to the secondary sepsis challenge. 32 These findings indicate that many not yet understood factors may regulate the responses of trained innate immune cells to subsequent challenges.
As neutrophils also exert a pathologic role in various inflammatory processes, such as thrombosis, atherosclerosis, periodontitis, or autoimmunity, 142,143 it is conceivable that maladaptive trained immunity may facilitate such pro-inflammatory and pathological functions of neutrophils in several contexts.
Nonmicrobial triggers are frequently linked with induction of maladaptive trained immunity in innate immune cells, including neutrophils. Tumors may hijack neutrophils to adopt a tumor growthpromoting phenotype. 144 Along this line, the in vivo exposure of mouse neutrophils to the microenvironment of mammary tumors increases, via a G-CSF-mediated mechanism, the in vitro and in vivo production of NETs upon re-stimulation with LPS, thereby promoting a pro-thrombotic effect. 145 Another important nonmicrobial training signal is hyperglycemia, the cardinal symptom of diabetes. Neutrophils isolated from blood of patients with type 1 or type 2 diabetes produce more NETs following ex vivo activation with phorbol-12-myristate-13-acetate (PMA) compared to those of healthy individuals and this is linked with the exposure of neutrophils to increased concentrations of glucose. 146 The hyperglycemiainduced training of neutrophils contributes to the delayed wound healing, a major clinical problem in diabetes. Indeed, diabetic mice display NET-dependent impaired wound healing capacity that is reversed upon deletion of PAD4, a calcium-dependent enzyme that is a key factor in mediating NETosis. 146 Based on the increased NET production of neutrophils when pre-exposed to hyperglycemia and the important role of NETs in the elimination of infectious microorganisms, one might hypothesize that hyperglycemia may have a beneficial role in the protection against infections. Nevertheless, this is not the case as neutrophils isolated from patients with type 2 diabetes produce more NETs (when activated in vitro with PMA or A23187) with diminished capacity to kill carbapenem-resistant and hypervirulent K. pneumoniae strain, compared to controls. 147 These findings potentially explain the increased susceptibility of type 2 diabetes patients to infection with such virulent bacteria. 147 The mechanisms underlying such maladaptive effects of hyperglycemia on neutrophils and whether they involve trained immunity-related epigenetic programs has not been elucidated yet and requires fur-  [151][152][153] Along the same line, western diet feeding of mice induces central trained immunity and increases the capacity of BM cells and splenic myeloid cells to respond to secondary stimuli, for example, to LPS ex vivo. 36 Western diet induces epigenetic and transcriptomic changes in GMPs toward a more activated phenotype; additionally, increased numbers and a more activated phenotype of monocytes and granulocytes are observed in the spleen of western diet-fed mice. The aforementioned effects of western diet are sustained even after switching mice to normal diet, clearly suggesting induction of detrimental trained immunity. 36 Feeding mice with a high-fat diet increases the numbers of LT-HSC, MPP, GMPs, as well as the potential for granulocyte and macrophage differentiation. Nevertheless, only increased levels of pro-inflammatory macrophages are detected following a switch to normal diet while neutrophils are unaffected. 154 As alluded to above, besides metabolic pathologies, inflammatory diseases, such as periodontitis, a dysbiosis-related inflammatory disease of the tooth-supporting tissues 155 or arthritis, may also induce detrimental central trained immunity involving neutrophils. 72 Periodontitis and associated bacteremias may cause systemic inflammation in humans with increased levels of inflammatory cytokines, such as IL-1 or IL-6, as well as elevated blood neutrophils, which display a hyper-responsive phenotype with increased generation of inflammatory mediators or ROS upon ex vivo re-stimulation. 100,[156][157][158] This hyper-responsive neutrophil phenotype may persist even after successful treatment of periodontal disease, thus indicative of innate immune memory. 159 Additionally, in humans periodontal inflammation positively correlates with BM hematopoietic activity, hence, suggesting increased myelopoiesis. 160 Interestingly, in patients with rheumatoid arthritis hematopoietic activity may remain higher in the BM despite clinical remission. 161 Figure 3). Intriguingly, BM neutrophils may also act as mediators of the periodontitis-induced central inflammatory memory. Specifically, enhanced gingival production of G-CSF due to periodontitis reaches the circulation and stimulates BM neutrophils to secrete IL-1β, which in turn acts on BM HSPCs to induce their training. 72 It becomes evident that the potential role of neutrophils as both effectors and mediators of beneficial or maladaptive innate immune memory induced by different microbial or host-derived triggers, including metabolic or inflammatory conditions, requires detailed mechanistic investigation in the future.

| FAC TOR S MODUL ATING THE OUTCOME OF TR AINED INNATE IMMUNIT Y
It is obvious from the aforementioned examples that trained immunity in general and neutrophil training in particular may be beneficial for the defense against infectious agents or against tumors but can also lead to increased inflammation and cardiometabolic diseases. 70 The factors that regulate this fine balance between beneficial and maladaptive inflammatory memory and neutrophil training remain largely unexplored. The strength of the training signal may play a critical role in the outcome of neutrophil training, as is implied by in F I G U R E 3 Maladaptive training of neutrophils in the BM contributes to the emergence of inflammatory comorbidities. Systemic inflammation induces epigenetically based inflammatory memory in HSPCs in the BM. Inflammation-adapted (trained) progenitors give rise to elevated production of neutrophils (and myeloid cells in general) with enhanced immune responsiveness to subsequent infectious or inflammatory challenges. These trained or hyper-responsive neutrophils infiltrate peripheral tissues, such as the periodontium and the joints, and exacerbate inflammatory disorders, such as periodontitis and arthritis. Because both periodontitis-induced and arthritisinduced systemic inflammation lead to inflammatory memory in the BM and increased susceptibility to either disease, it is concluded that maladaptive innate immune training of BM hematopoietic progenitors links distinct inflammatory comorbidities. GMP, granulocytemonocyte progenitor; HSC, hematopoietic stem cell; MPP, multipotent progenitor.
vitro studies inducing training of mouse BM neutrophils with a big range of LPS concentrations, 162 with the component of gram-positive bacteria lipoteichoic acid 163 and with isolated small gut microbiotaderived extracellular vesicles. 164 The aforementioned studies demonstrate that exposure of neutrophils to low concentrations of the training agent enhances the production of pro-inflammatory molecules, such as TNF, IL-6, MCP-1 or IL-1β, generation of ROS, and their phagocytic and migratory capacity upon restimulation, while high concentrations of the training agent result in increased production of anti-inflammatory agents, such as IL-10, as well as impaired phagocytosis and migration in vitro. [162][163][164] The route of administration of the training factor may also determine the subsequent responses and outcomes. For instance, intravenous administration of BCG vaccine to mice directs the accumulation of BCG in the BM and promotes expansion of HSPCs, while subcutaneous administration of the same vaccine does not affect HSPC numbers. 75 Interestingly, the duration of the training effect induced by a COVID-19 infection was proportional to the disease severity. 141 In humans, the training outcomes may be modulated by genetic predisposition as well. Polymorphisms in HNF1A, HNF1B, GATA2/3, GFI1/1B, HOXA4, and visual system homeobox 1 (VSX1) alter the responses of previously BCG or β-glucan ex vivo trained monocytes to a subsequent LPS-stimulated production of IL-6 and TNF. A detailed analysis for the identification of predictive biomarkers that could forecast the efficiency of the innate immune training response was also conducted. The elevated plasma levels of CCL20 and CCL23 correlated with the elevated production of IL-1β and IL-6 by BCG vaccine-trained peripheral blood mononuclear cells re-exposed in vitro to C. albicans. 125

| INNATE IMMUNE TR AINING INDUCER S IN THER APEUTI C APPLI C ATI ON AG AIN S T C AN CER
The training agonist BCG is used as an adjuvant therapy for bladder cancer for over 40 years now 165 ; BCG immunotherapy is successfully used for the therapy of non-muscle invasive bladder cancer. 166,167 The therapeutic effect of BCG is attributed, at least in parts, to neutrophils since their deletion abrogates the BCG-mediated anti-tumor function in mice. 168 Neutrophils are the first cells that infiltrate the bladder after BCG treatment where they produce high amounts of chemokines and cytokines and direct trafficking of T cells and other immune cells to the bladder. 168,169 Moreover, human and mouse neutrophils exhibit direct tumor killing capacity by releasing increased amounts of TRAIL and upregulate NETs production after BCG treatment. 170,171 β-glucans are also used for cancer treatment in experimental models for decades now, long before their characterization as innate immune training agonists. 172 In most cases, β-glucan-based regiments are currently combined with other immunotherapies, anti-angiogenic or tumor cell-targeting therapeutics. 173 Based on the success of such combinatorial approaches in syngeneic mouse tumor models 174,175 and human xenograft models, 176,177 clinical trials are underway using β-glucan in combination with further anti-cancer therapies, such as immunotherapies or anti-angiogenic antibodies targeting vascular endothelial growth factor. 178,179 Neutrophils play a central role in anti-tumor-immunity induced by the β-glucanrelated therapies. 37,180,181 For instance, adoptive transfer of neutrophils from β-glucan-trained mice inhibits tumor growth, suggesting that β-glucan-trained neutrophils may represent an adjuvant cancer immunotherapy. 37 187 can trained immunity affect tissue repair processes? Can trained immunity affect tissue homeostasis? What mechanisms underlie the development of protective versus maladaptive trained immunity? Ostensibly, microbial products may induce trained immunity in neutrophils associated with protective and beneficial outcomes, while sustained chronic inflammation seemingly promotes maladaptive inflammatory memory in granulopoiesis. 37,72,124 In this regard, the conceivable question is whether the duration and strength of the training signal determine the training outcome in a quantitative and qualitative fashion. Delineating the factors, variables, and mechanisms that favor the protective versus the detrimental actions of neutrophilrelated trained immunity are of great importance for the design of potential future therapeutic interventions.

ACK N OWLED G EM ENTS
We apologize to authors whose work could not be referenced

CO N FLI C T O F I NTE R E S T
The authors have no conflict to disclose.

DATA AVA I L A B I L I T Y S TAT E M E N T
There are no data in this manuscript.