Re ‐ recognition of innate immune memory as an integrated multidimensional concept

In the past decade, the concept of immunological memory, which has long been considered a phenomenon observed in the adaptive immunity of vertebrates, has been extended to the innate immune system of various organisms. This de novo immunological memory is mainly called “ innate immune memory ” , “ immune priming ” , or “ trained immunity ” and has received increased attention because of its potential for clinical and agricultural applications. However, research on di ﬀ erent species, especially invertebrates and vertebrates, has caused controversy regarding this concept. Here we discuss the current studies focusing on this immunological memory and summarize several mechanisms underlying it. We propose “ innate immune memory ” as a multidimensional concept as an integration between the seemingly di ﬀ erent immunological phenomena.


INTRODUCTION
Traditionally, the host immune system has been divided into innate and adaptive immunity. Innate immunity is the primitive immune system that is conserved among multicellular organisms, whereas adaptive immunity has evolved in the vertebrate lineage. The innate immune system detects pathogen-associated molecular patterns (PAMPs) via pattern recognition receptors (PRRs). 1,2 This simple recognition system is rapidly activated by a broad range of pathogens, resulting in a transient and nonspecific response. 3 This system relies on macrophages, natural killer (NK) cells, and dendritic cells (DCs) in mammals and analogous cells in other organisms. On the other hand, adaptive immunity exhibits a high specificity against pathogens through antibody-antigen reactions, and in mammals it is mainly dependent on two types of lymphocytes (B and T cells). 4 In contrast to the innate immune response, the lymphocytedependent adaptive response is slower but generates long-term protection, defined as immune memory. 5 The lymphocytes that undergo genomic recombination specifically recognize pathogens and are maintained as memory cells for future immune responses. It has long been considered that innate immunity does not have such a memory function. However, a growing body of evidence has suggested that innate immunity also forms immunological memory. Innate immune memory has a stronger recall and faster response to subsequent infection. Memory features have been evaluated through traits such as increased survival, 6 reduced bacterial load, 7 and differential gene expression. 7,8 The innate immune memory phenomenon has also been detected in vertebrates, plants and invertebrates. [9][10][11] Understanding the characteristics of this immune memory may potentially lead to new approaches for the development of vaccines and disease therapies. Many studies are currently investigating the mechanisms underlying this innate immune memory.

INNATE IMMUNE MEMORY AMONG DIFFERENT SPECIES
The first evidence of the innate immune system forming memory features was observed in a study of plants in 1933. It was reported that in plants a local pathogen infection induced greater protection against reinfection of the whole body. 12 Later, it was also shown that inoculation with tobacco mosaic virus (TMV) on half-leaves of Nicotiana tabacum caused resistance to TMV and other viruses even on the opposite half-leaves. 13 This type of plant immunity is termed "systemic acquired resistance" (SAR). 14 SAR is able to protect plants from numerous pathogens in a broad-spectrum manner, and this protection can last from a few days to lifelong. 9,15,16 Thus, SAR is considered a type of innate immune memory in plants.
In addition, it has long been known that prior pathogenic exposure induces protection against subsequent infection in invertebrates which lack adaptive immunity. Kurtz et al. investigated "immune priming" using the copepod Macrocyclops albidus against its natural parasite, the tapeworm Schistocephalus solidus. 17 M. albidus was exposed to S. solidus for the first infection, and three days later they were challenged with the equivalent tapeworm or unrelated parasites. They found that the equivalent tapeworms had less success and intensity of infection after the subsequent challenge than the unrelated parasites, which suggests the presence of innate immune memory. Moreover, priming with a low-dose infection of Streptococcus pneumoniae increased the survival rate of fruit flies (Drosophila melanogaster) against a subsequent challenge with a lethal dose of S. pneumoniae. 6 Another study showed that snails, Biomphalaria glabrata, primed with its compatible sympatric parasite, Schistosoma mansoni, completely resisted infection when challenged a second time with the same parasite. 18 In addition, immune memory has been broadly observed in numerous different invertebrate species such as the silk moth Bombyx mori, roundworm Caenorhabditis elegans, and mosquito Anopheles gambiae. [19][20][21] Lanz-Mendoza et al. 22 summarized the studies of invertebrate immune memory in a comprehensive review.
The enhancement of the innate immune response at a secondary challenge has also been observed in mammals and this phenomenon has been termed "trained immunity". 11,23 This concept has been described as the long-term memory in innate immune cells, which is evoked by pathogen stimuli and leads to altered responses against secondary heterologous or homologous challenges. 24 Here the nonspecific protection of Bacillus Calmette-Guérin (BCG) vaccine is given as a typical example. Initially, BCG was used to protect infants from infection with Mycobacterium tuberculosis. However, it has been proven that BCG also protects against various heterologous pathogen infections. [25][26][27] Since the BCG-induced immune response was identified not only in lymphocytes but also in innate immune cells such as monocytes, 28,29 the nonspecific beneficial effects of BCG were thought to be produced by innate immune memory. Furthermore, studies using a severe combined immunodeficiency mouse model, that lack an adaptive immune response, revealed an enhanced immune response at secondary infection, which suggests the existence of memory features in mammal innate immunity. 30 Thus, the phenomenon of innate immune memory seems to be evolutionarily preserved among the different species from plants to vertebrate metazoans.

TERMINOLOGY: HOW TO DEFINE INNATE IMMUNE MEMORY
That the innate immune system in many organisms possesses memory characteristics is widely accepted. However, because the mechanism underlying this is unclear, the concepts and terminology of this phenomenon remain controversial. In some invertebrates, the memory-like property of innate immunity is defined as the specific protection against a secondary challenge by the same pathogens. 22,31 This has been termed "immune priming" as a specific and long-lasting innate immune memory, [31][32][33] whereas "immune enhancement" was defined as nonspecific responses against challenges by other pathogens. 22,34 Under some circumstances, homologous challenges (a secondary infection by a pathogen which is related to the primary pathogen) are required to test the innate immune memory, whereas heterologous challenges (a secondary infection by a pathogen which is unrelated to the primary pathogen) is sufficient to test the immune enhancement. 22 In other cases, especially in vertebrates, the adaptive characteristics of innate immune responses are often termed "trained immunity". Trained immunity is used to describe the enhanced immune response after a primary infection dose (training), and the protective effects are usually nonspecific. 25,26,30 Thus, "immune priming" is used to describe the homologous long-lasting memory feature in invertebrates, whereas "trained immunity" and "immune enhancement" are used to describe heterologous immune responses in vertebrates and invertebrates, respectively. However, some studies use these terms for different meanings. [35][36][37][38][39][40][41] Inconsistent definitions have produced ambiguities that prevent researchers from using the appropriate terminology.
In this review, we use these terms to describe each cited study, but also "innate immune memory" as a broad term that incorporates the memory phenomenon of the innate immune system in all phyla. We consider that innate immune memory should be considered as a multidimensional notion. Several major dimensions of innate immune memory are described below.

Specificity
In general, specificity is the dimension indicating whether the secondary stimulus is related to the primary stimulus. As mentioned, some studies have claimed that "memory" should be specific, whereas others have regarded memory as nonspecific stimulation. In some cases, innate immune responses such as the antiviral RNAi system in eukaryotes or the CRISPR system in bacteria and archaea could exhibit high specificity similar to that of adaptive immunity. However, in other cases, innate immunity shows broad-spectrum specificity. 24 Adaptive immunity depends on one-to-one antibody-antigen responses, whereas innate immunity recognizes pathogens through sensing PAMPs by PRRs. 1,2 Since PAMPs are conserved within a group of pathogens (e.g., lipopolysaccharide, LPS, in gram-negative bacteria), innate immunity recognizes a group of pathogens instead of the specific pathogen itself. This may be a reason why innate immune memory appears to be specific or nonspecific, depending on the experimental design. An interesting opinion using the philosophical concept of "multiple realization" predicts that immune memory can be implemented through different pathways. 42 Therefore, it is unsurprising that different species utilize different mechanisms for the same purpose of immunological memory. In other words, it is possible that immunological memory against a given pathogen could exhibit specific protection against the same pathogen in some species and broadspectrum protection in others. Moreover, even within a species, different mechanisms might be integrated and exhibit immunological memory in a specific or nonspecific manner depending on the pathogen.

Memory or persistence
It is generally accepted that memory refers to a response that is recalled after the clearance of the primary stimulus. Therefore, to assess "memory", the immune response should return to a baseline before the secondary infection 43 (Figure 1a). Otherwise, the immune response would be maintained through continuous pathogenic stimulation (Figure 1b), and this phenomenon is termed "persistence" rather than "memory". A pathogen load assay and transcriptomic analysis can be used to assess whether the stimulated response results from memory or persistence. The pathogen load assay evaluates whether the pathogen of the primary infection persisted until the secondary challenge. Transcriptomic analyses reveal patterns of gene expression that are indicators of the immune phase. For example, our recent RNA sequencing (RNA-Seq) study provided an unbiased validation of the persistence and memory immune responses in D. melanogaster. 7 Training with primary infections of Salmonella typhimurium (St) and Micrococcus luteus (Ml) stimulated immune responses during the secondary infection challenge in flies. However, we found that St bacteria persisted in the flies for a prolonged duration, whereas Ml bacteria were eliminated after day 12 after training infection. Furthermore, RNA-Seq analysis revealed that the immune response gene expression persisted after St training but returned to baseline 6 days after Ml training. However, after the secondary infection challenge, the Ml training stimulated the immune response gene expression. These results suggest that St and Ml infections induce "persistence" and "memory" immune responses, respectively. Thus, future studies should focus on evaluating this dimension of innate immune memory.

Duration
Memory duration varies from a short period (about a few days) to long-lasting protection (up to the next generation) across different phyla. In invertebrates, evidence has shown that immune memory lasts for up to a few weeks in some cases 22,[44][45][46] and is capable of lasting from the larval to the adult period in others. 21 According to our recent study, survival increased as the memory effect was maintained on day 12 after Ml training, when Ml bacteria were eliminated. 7 A period of 12 days is about one-quarter of an adult fly's lifespan, which suggests that innate immune memory might provide significant advantages in nature for the species. As a further example, the priming effect in invertebrates has been passed maternally or paternally to the next generation, thus stimulating pathogen resistance in the offspring. This memory-like property is particularly known as transgenerational immune priming (TGIP). 47 In vertebrates, research is primarily conducted using cultured cells and the duration of innate immune memory could be diverse across the different cell types ranging from days to several months or years. 24 Initially, trained immunity was demonstrated in mature myeloid cells such as DCs, monocytes, and NK cells. [48][49][50] For instance, BCG vaccination enhances the proinflammatory interferon production in NK cells upon ex vivo infection. This heterologous beneficial effect persists for at least 3 months after BCG vaccination. 51 However, considering that mature myeloid cells have a short lifespan of about 4-7 days in vivo, 52 it was puzzling how immune memory was maintained over the long-term. In recent years, several studies have shown that bone marrow cells such as hematopoietic stem cells (HSCs) are capable of building trained immunity. HSCs can renew themselves and have a long lifespan of up to 60 months. 53 HSCs respond directly to infections and produce bone marrow-derived mature cells with immune memory. Netea et al. reviewed the central and peripheral trained immunity in detail. 24 It is worth mentioning that transgenerational immune memory is observed in both vertebrates and invertebrates. In this condition, male mice trained with Candida albicans conferred heterologous immune stimulation to the next generation. 54 There is no doubt that the phenomenon of immune memory varies among species and therefore, it is challenging to discuss all these phenomena together. It is feasible to adopt the diversity of new immunological memory as an integrated multidimensional notion.

MECHANISMS
The crucial role of innate immune memory properties has been recognized and has been focused on in recent years in both invertebrates and vertebrates. As a multidimensional phenomenon, different mechanisms contribute to the response. Although the comprehensive mechanism that underlies this process remains unclear, several clues have been observed over the years.

Epigenesis
Epigenetic reprogramming has long been proposed as an important player in the innate immune memory process. Epigenetic reprogramming occurs through changes in the chromatin reorganization (e.g., DNA modification and histone posttranslational modification) and thereby changes the gene expression. 55 When a host is infected with a pathogen, immune signals change the gene expression patterns. 24 There is plenty of evidence suggesting that innate immune cells, such as monocytes and macrophages in vertebrates 29,56 or hemocytes (macrophages-like cells) in invertebrates 57,58 have the capability to leave "epigenetic marks" within some genes. The epigenetic marks change the chromatin structures from the condensed to the open state, thus altering the accessibility of DNA to transcription factors. After primary infection, epigenetic marks remain for a long time, and the open chromatin states are ready to facilitate gene expression under subsequent challenge conditions. 24 For example, many studies have reported that mammalian macrophages and NK cells stimulated with LPS and other PAMPs resulted in an increased level of trimethylation at histone 3 lysine 4 (H3K4me3) and acetylation at histone 3 lysine 27 (H3K27ac) at the inflammatory gene loci, thus leading to altered expression at secondary stimulation. 24,29,43,59 In addition, our recent RNA-Seq study identified Ada2b as a factor involved in innate immune memory of the fruit fly. 7 Ada2b is a component of the Spt-Ada-Gcn5 acetyltransferase (SAGA) complex and interacts with the histone acetyltransferase (HAT) module to regulate the acetylation of histone 3 (H3K9ac and H3K14ac). [60][61][62] The SAGA complex also contains the DUB module that regulates histone deubiquitylation. 62 However, it is unknown which histone modifications involving the SAGA complex contribute to innate immune memory. Moreover, various histone modifications (e.g., ubiquitination, phosphorylation, acetylation, and methylation) lead to the induction and inhibition of gene expression. [63][64][65] Therefore, it is possible that different histone modifications of different genes can regulate the gene expression in either a positive or negative direction in the memory process. However, it is still unclear how these epigenetic marks are written at specific genome locations and how they enhance or suppress the gene expression. Understanding these issues could be pivotal and should be considered in future research.

Metabolism pathways
The rewiring of intracellular metabolic pathways is another hallmark involved in the innate immune memory process. Cells change their bioenergetic states for innate immune memory and adaptation to nutrient conditions. Glycolysis is a major energy producing metabolic pathway in cells and is considered crucial in the development of innate immune memory. For example, in vertebrate myeloid cell lineages, inflammatory signals induce aerobic glycolysis that can rapidly produce adenosine triphosphate (ATP) and pyruvate. Pyruvate is then oxidized to lactate or enters the TCA cycle to produce more ATP. A metabolic shift from oxidative phosphorylation toward aerobic glycolysis is observed in immune cells stimulated with β-glucan and BCG, which induce innate immune memory. [66][67][68] Notably, inhibiting the mTOR pathway using metformin decreased the cytokine levels in response to the restimulation of both β-glucanand BCG-trained immune cells ex vivo, suggesting that the AKT-mTOR-HIF1α pathway controls the metabolic pathway involved in the innate immune memory process ( Figure 2). 69,70 In recent years, a growing body of evidence has indicated that the western diet and obesity are risk factors for the inappropriate induction of immune memory in myeloid cells, suggesting a role of lipid metabolism in immune memory. [71][72][73] The western diet usually consists of fast foods, snacks, and sweets, which are abundant in sugars and animal fats while lacking in fiber and vitamins. 74 High levels of fatty acids are sensed as danger signals by PRRs such as the Nod-like receptor family pyrin domain containing 3 (NLRP3) and toll-like receptors (TLRs), which stimulate inflammatory responses against further stimuli. 71,75 Accordingly, oxidized low-density lipoprotein (oxLDL) enhances glycolytic activity and induces innate immune memory in macrophages, thus suggesting the essential role of glycolysis and lipid synthesis pathways in forming the immune memory phenotype. 73,76,77 Apart from these, the involvement of other metabolic pathways (e.g., pentose phosphate pathways and amino acid metabolic pathways) in innate immune memory has also been reviewed in detail. 78,79 It is still poorly understood how cellular metabolism connects to epigenetic memory in innate immunity, although it is suggested that some products of the TCA cycle (e.g., acetyl-CoA) might be involved in histone modifications during immune memory (Figure 2). 80 Our recent study estimated the genomic variations of innate immune memory in Drosophila and identified several genomic loci that potentially correlate to the variations. 81 Interestingly, one of the candidate genes, Adgf-A (adenosine deaminase growth-factor A), was involved in the memory process.
Since Adgf-A encodes a degradation enzyme of extracellular adenosine, our results suggest that the adenosine (Ado) pathway might contribute to immune memory. We found that AdoR (Adenosine receptor), as well as Adgf-A, was required for innate immune memory (unpublished data). Ado is a crucial signaling molecule for both immunity and metabolism, and is known to regulate systemic inflammatory responses and metabolic changes in both invertebrates and vertebrates. [82][83][84] Ado can be transported in and out of cells by the transporter, however, it is unclear how intracellular and extracellular Ado is regulated in immune responses. The key source of intracellular Ado is the S-adenosyl-l-homocysteine (SAH), produced from S-adenosyl-l-methionine (SAM), 85,86 and SAM is considered a methyl donor for DNA and histone methylation. 87 Consistent with this, increased methionine consumption was detected in monocytes exposed to β-glucan as a memory-related phenotype. 41 Furthermore, Ado in endothelial cells induce a hypomethylated state of DNA at the promoter region of angiogenetic genes in a HIF1αdependent manner, which suggests the potential role of Ado in epigenetic reprogramming ( Figure 2). 88 Thus, the possibility that the methionine-SAM-Ado metabolic pathway links cellular metabolism and epigenetic reprogramming should be considered. Systemic metabolic rewiring and epigenetic reprogramming may be highly integrated in immune responses and cause memory status for future infection.

Noncoding RNAs
Organisms have developed immune mechanisms against viruses. The RNA interference (RNAi) system is known as a protection system against viruses in eukaryotes including F I G U R E 2 Metabolism pathways and epigenetic reprogramming during immune memory. The primary infection triggers the activation of intracellular metabolism pathways such as glycolysis and adenosine metabolism. Some metabolites of these processes, such as acetyl coenzyme A (acetyl-CoA) and S-adenosylmethionine (SAM), might mediate epigenetic reprogramming of gene regulation during innate immune memory.

RE-RECOGNITION OF INNATE IMMUNE MEMORY
plants, invertebrates, and vertebrates. Since the late 1990s, various types of small RNAs (e.g., siRNAs, miRNAs, and piRNAs) were discovered, and their roles in regulating gene expression and virus resistance have been revealed. [89][90][91][92] In Drosophila, it was reported that hemocytes incorporate viruses through phagocytosis and produce virus-derived siRNAs for antiviral immunity. 93 In response to exogenous RNAs, Dicer family enzymes are activated and generate miRNAs and siRNAs with complementary sequences as exogenous RNA molecules. These small RNAs are loaded into RNA-induced silencing complexes under the function of Argonaute (AGO) subfamily proteins and directly cleave virus RNAs. 94,95 Tessetto et al. demonstrated that siRNAs produced by hemocytes were subsequently incorporated into extracellular vesicles (EVs), such as exosomes, and EVs are distributed to transduce their systemic antiviral signals to other naïve cells. 93 Furthermore, TGIP was demonstrated in the RNAi system of the fruit fly. Once priming virgin female flies with viruses such as Sindbis virus (SINV) and Drosophila C virus, virus-derived DNA (vDNA) was detected until the F2 generation, and these F2 individuals resisted virus infection. 96 Although it is unclear how vDNA are produced in the subsequent generations, it is possible that siRNAs might act as a molecule that transmits memory information. Moreover, small RNAs are not only vertically transferred from parent to offspring, but can even be horizontally transferred between individuals. 97,98 Similar phenomena were observed in vertebrates where EVs carrying virus-derived siRNAs were detected in mice blood when infected with various viruses. 99,100 These results suggest that EV-dependent siRNA transmission is an evolutionarily conserved mechanism for immune memory in subsequent generations. Another antiviral immune system exists in bacteria and archaea, and this system is capable of memorizing and preventing virus infection in an adaptive manner, termed "Clustered Regularly Interspaced Short Palindromic Repeats and the associated proteins (CRISPR-Cas)". 101 CRISPR is a series of DNA in the bacteria and archaea genome and originates from previously infecting bacteriophage (virus) DNAs. 102 During phage infection, Cas complexes recognize exogenic DNA and cleave it into small pieces (spacers), and these pieces are integrated into a repeat-spacer array, forming the CRISPR locus. When exposed to a secondary infection, the CRISPR locus is transcribed into a long pre-CRISPR RNA (pre-crRNA) molecule and trimmed into smaller mature CRISPR RNAs (crRNA) by ribonucleases. Subsequently, the crRNA specifically guides the Cas proteins to the foreign nucleic acids through base-pairing interactions, and the Cas proteins degrade the foreign nucleic acids. 101,103 The CRISPR-Cas system provides an adaptive immunity-like response and is highly conserved in about 50% of bacteria and 90% of archaea. 104 Although both RNAi and CRISPR-Cas are well known as powerful technologies in molecular genetics, their functions are related to innate immune memory. It is possible that understanding the mechanisms of innate immune memory could contribute to the improvement of these technologies and the development of new ones.

CONCLUSION AND DISCUSSION
In this review, we discuss some important recent and current studies on innate immune memory. The notion of immunological memory has been expanded to include innate and adaptive immunity over the past decades. Although the distinction between innate and adaptive immunity is still a useful framework to understand the different branches of the immune system, the boundary between them is not always clear. For example, some components of the innate immune system, such as DCs, present antigens to the adaptive immune cells, leading to more effective responses in subsequent infections. This process known as "cross-presentation", can provide the innate immune system a degree of specificity and memory-like features. 105 On the other hand, some components of the adaptive immune system, such as CD8 + T cells, undergo cytokine-mediated activation in nonspecific immune responses. 106 Another example is the CRISPR-Cas system. The CRISPR-Cas system uses pathogen DNA as the "memory" of infection, thereby providing a rapid and specific defense against the pathogen. Therefore, memory is a feature overlapping between innate and adaptive immunity.
Innate immune memory is attracting increasing attention and offers potential therapies for several diseases. For instance, research has developed a new type of cancer immunotherapy that activates trained immunity using bioengineered nanomaterial. This therapy could be applied even under immunosuppressive conditions to enhance the antitumor effects of myeloid cells. 107 In addition, recent studies have shown that the BCG vaccine induces innate immune memory in humans and is a potentially effective prophylactic treatment for COVID-19, 108,109 although its effects are still controversial. 110,111 Furthermore, immunological memory is also applicable to other organisms. For example, the first TGIP vaccine in invertebrates was used in honeybees to reduce their mortality from the infection with Paenibacillus larvae. 112,113 Extending the concept of immunological memory provides various potential applications, not only in vaccine manufacturing or therapeutic approaches for humans, but also in agriculture and stockbreeding if we could develop vaccines for vertebrates, invertebrates, and plants.
Adaptive immunity is the defense system specific to vertebrates, whereas innate immunity is the primitive immune system shared by diverse organisms. Since innate immune memory is based on various molecular mechanisms, including epigenetics and cellular metabolism, it may have evolved originally or independently in the lineages of organisms. It would be interesting and beneficial to determine which mechanisms of innate immune memory are conserved among organisms or which are specific to each lineage. Such conservation may shed light on the evolutionary history of the primitive immune system.