Immune memory is a well-studied phenomenon in mammals, and forms the basis for immune protection triggered by previous infections or vaccination. Exposure to microbial agents or their components can elicit an antigen-specific adaptive immune reaction targeting the triggering agent, encompassing production of neutralizing and opsonizing antibodies, generation of CD4⁺ and CD8⁺ cytotoxic T cells and other mechanisms (1). Typically, this response requires some time for developing (7–10 days), as the adaptive immune system should generate, by somatic recombination, specific tools uniquely recognizing and blocking the foreign agent (e.g., antibodies). Such adaptive immune response can be active for a period of time, as required for eliminating the danger, but it progressively wanes once the triggering agent is eliminated. This has become clear to everybody during the recent SARS-CoV-2 pandemic, as people received several vaccine jabs every few months, in order to sustain the anti-viral antibody levels, which otherwise would have disappeared. The protective immunity obtained by sustaining the immune response with repeated exposures to the antigen aims at achieving immediate inactivation of the infectious agent, when the pathogen is very rapidly infective and life-threatening.

Immune memory is a protective mechanism different from such sustained response (2). In mammals, upon elimination of the infective agent, the organism stops its immune activation (which is highly energy-demanding), and the immune system returns to baseline quiescence. However, memory cells generated by the encounter with the pathogen persist in certain niches in the body, and are already equipped to produce the required antigen-specific tools if/when the pathogen will come again (3, 4) Upon re-infection, such memory cells can mount a more efficient antigen-specific response in a much shorter time (1–2 days). The protective efficacy of immune memory depends on the type of pathogen. Infections with an incubation period of few days can be tackled effectively by immune memory cells, whereas pathogens that produce immediate damage (e.g., by producing toxins) can only be neutralized by a sustained response (e.g., pre-formed antibodies). Protective immunity and immune memory, both in response to infections and upon vaccination, are however more complex than a pathogen-specific response as described above, pertaining to the vertebrate/mammalian adaptive immunity.

The ancient evolutionarily conserved immune mechanisms of innate immunity are also primarily engaged in pathogen identification and elimination (5). Innate immunity is based on germline-encoded tools, such as the TLR receptors, and, because of this characteristic, is immediately ready to act against pathogens. For this reason, it was considered unable to “learn” and therefore to develop memory (6, 7). However, from evidence in plants and invertebrate metazoans, which only display germline-based immunity, as well as mammals, it is now clear that innate immunity can develop memory-like responses, based on different mechanisms and tools (e.g., metabolic and epigenetic changes that modulate and select subsequent gene expression, rather than somatic recombination that produces new effector molecules) (8–16). This memory-like immune reactivity is a mechanism of adaptation to the environment-derived cues (e.g., infectious agents), but it can be considered memory (which by the name must be specific and long lived) (2) because it can persist for the entire organism lifespan and be transmitted to the progeny, and it can also display a significant degree of specificity (see below). On this basis, we will define it as innate immune memory in the following paragraphs. The capacity of the innate immune memory responses to confer specific protection against pathogens is still a matter of investigation. Indeed, the dogma that innate immunity is fully non-specific (based on the lack of specificity of its recognition tools) is challenged by the abundant evidence in invertebrates and by accumulating evidence in mammals as well. We will discuss such evidence and propose a hypothesis on why and how innate immunity and its memory need and attain some degree of specificity despite its non-specific recognition tools.

Immune memory should be considered as a multidimensional concept, which is then declined differently in different organisms and with different underlying mechanisms (e.g., germline-encoded tools and epigenetic changes vs. somatic recombination mechanisms). The five key dimensions, as described by Pradeu and Du Pasquier (16), are: 1. strength, 2. duration, 3. speed, 4. specificity, and 5. extinction. Somatic recombination-dependent immune memory is in general characterized by high strength in the secondary response, a long duration, enhanced speed in the secondary response, high specificity and a clear extinction phase between the first and the second exposure. In organisms devoid of adaptive immunity (in particular in plants and invertebrate metazoans) the generation of memory-like immune adaptation is a well-known phenomenon, with organisms able to mount a more efficient immune response when re-exposed to a challenge (usually an infectious agent) (8, 9, 11, 15, 17). Memory-like immune reactivity in invertebrates has been described in at least three different forms, the recall response, the immune shift and the unique sustained response/acquired resistance (9, 16, 18–21). The recall response implies extinction of the immune response after the first exposure and a more effective response upon subsequent exposure. Notably, there is no evidence of an enhanced speed, most likely due to the fact that innate immunity is already very fast at the first challenge. This kind of memory-like response reflects the kinetic profile of adaptive immune memory in mammals, such as that induced by vaccination (22). The immune shift implies, after the first response, the generation upon re-exposure of a different and more protective kind of response, e.g., from phagocytosis to encapsulation (23, 24). It is interesting to note that also in this case there are similarities with adaptive immune memory, in which re-exposure induces a maturation of response, such as for instance the immunoglobulin class switch towards more efficient antibody classes (25, 26). The unique sustained response/acquired protection is probably not a memory response sensu stricto, since there is no extinction of the immune activation induced by the first exposure. Here, the response to a first challenge is kept at increasingly higher levels by re-exposure, achieving improved protection (27). The unique sustained response reflects the protective adaptive immunity attained in human beings by re-stimulation, e.g., with repeated vaccine boosts (28). The persistence of the antigen/agent that has triggered the response, even at low concentrations, is probably one of the elements at the basis of the persistent unique response in invertebrates, as well as of the sustained protective immunity in mammals. In this sense, the unique sustained response can be considered as an highly effective adaptation behavior, rather than a true memory response (2). However, since memory is linked to epigenetic reprogramming (both in animals and plants) (17, 29), the duration of immune memory is expected to depend on the persistence of the epigenetic changes underlying it, which in turn may depend on the presence of the memory-inducing agents/microorganisms in the living environment (30). In any case, other mechanisms may intervene in the establishment and maintenance of memory, including germline DNA modifications (see below). Thus, immune memory can last for the entire life of the organism, be maintained in daughter cells and transmitted to the progeny (9).

Specificity of the memory-like responses in invertebrates has been shown in several studies, mostly based on infection and re-infection of animals and assessment of the extent of secondary responses towards the original vs. unrelated microorganisms (31–40), although some studies might be biased by inadequate experimental design (16). Depending on the specificity of the secondary response, the invertebrate immune memory has been defined as immune enhancement (non-specific increase of secondary protective immunity) and immune priming (specific induction of improved secondary protective immunity) (see definitions in 20). The few examples below are meant to illustrate some of the many scenarios of specificity vs. selectivity of invertebrate immune memory.

In Porifera (sponges) and Cnidaria (corals), immune memory was described since the ‘80ies of last century as capacity of self/non-self-recognition in transplantation and parabiosis experiments (20, 41), leading to an accelerated specific rejection/destruction of a second allograft (42, 43). The rejection needs cell contact and is likely mediated by macrophage-like amoebocytic cells (41). Exposure of Drosophila to a sublethal dose of Streptococcus pneumoniae could induce a phagocyte-dependent life-long specific protection against a subsequent lethal challenge with the same microbe (36). Previous exposure of the oyster Crassostrea gigas to inactivated Vibrio splendidus bacteria could induce a potent secondary response to live V. splendidus, in terms of hemocyte recruitment and phagocytosis, but not to other Vibrio ssp. or unrelated bacteria (40). This memory-like immune adaptive response was at the level of the entire organism, as it depended on the specifically enhanced influx of hemocytes, while re-programming of cellular functions did not occur, as the phagocytic ability of individual hemocytes did not change. Conversely, in Anopheles gambiae, infection with Plasmodium berghei induced a protective memory recall response to P. berghei and also P. falciparum, which depended on the recruitment of the granulocytic effector cells to the infections site and their reprogramming towards enhanced parasite elimination (44). Another study in the terrestrial arthropod Porcellio scaber shows that in vivo infection with Bacillus thuringiensis strain 1 induced the subsequent increase of hemocyte phagocytic capacity against B. thuringiensis 1 and 2, but not Escherichia coli, a finding that suggests a re-programming at the cellular level (33). Hemocytes from the shrimp Litopenaeus vannamei exposed to the inactivated pathogen V. harveyi can develop an increased capacity to phagocytose the same bacteria, but not the unrelated Bacillus subtilis, upon a subsequent exposure (34). However, priming with B. subtilis could not induce a specific secondary increased phagocytosis, suggesting that memory induction was selective. In larvae of Bombyx mori, exposure to different bacteria raised a specific secondary phagocytic response, which could distinguish between Gram-positive and Gram-negative bacteria and also different strains within the same Gram-type. The memory response was both at the organism level (increased number of phagocytes and increased production of antibacterial factors) and at the cellular level (increased phagocytic capacity of individual cells). The memory response was also specifically protective against infection with live bacteria (35). Interestingly, an immune protective memory recall response was induced in the planaria Schmidtea mediterranea by exposure to the natural pathogen Staphylococcus aureus, but not by infection with Legionella pneumophila or Mycobacterium avium, which are not natural pathogens. The recall response was based on epigenetic changes and involvement of stem cells, which did not occur upon priming with non-natural pathogens (45). In the ascidian Ciona robusta, priming with gram-negative LPS or gram-positive LTA induced a memory recall response that largely depended on the secondary challenge. Irrespective of the priming agent, the recall response induced by an LPS challenge was a more potent cellular response (increased phagocytosis and expression of cellular receptors), while the recall response to LTA was a more potent humoral response (decreased phagocytosis, increase expression of complement components and cytokines) (19). In the crustacean Artemia franciscana, an abiotic stress (heat shock) could induce a protective secondary response both against subsequent heat shock and against infection with pathogenic V. campbellii (46). In C. robusta, exposure to abiotic stress (hypoxia + starvation) induces a recall response to bacterial LPS that differs between immune-competent organs: a general downregulation of cellular and humoral response genes in the pharynx and a general upregulation in the gut (47).

Thus, the immune memory-like adaptation responses in invertebrates can occur in different forms and adopt different degrees of recognition and specificity, but all aiming at affording better protection against recurrent pathogens or abiotic stress present in their living environment. Thus, a fully specific memory can be generated, for the targeted recognition and resistance to recurrent pathogens, which most likely are constantly present in the environment and provide the necessary level of re-boosting to afford a unique sustained response/acquired protection (33–35, 40). Also, memory seems to develop in response to natural pathogens but not upon priming with non-natural microorganisms (33, 44). The memory response can be challenge-specific, with priming just non-specifically making the immune system “ready” to adapt its memory response to the type of challenge (19). Eventually, the memory response induced by abiotic stress can be of broad specificity, affording a protective response against the specific agent but also against natural pathogens (46), and it can substantially differ in different immune-competent organs (47). The overall goal all these different immune memory-like forms is the successful adaptation to the environmental challenges, by developing a more protective response against recurrent specific threats.

While the mechanisms of memory responses are based on the same mechanisms of primary reactions (e.g., engagement of TLR and other innate receptors, RNA interference, phagocytosis, production of toxic molecules, etc.), the mechanisms by which memory is established and transmitted are not fully elucidated and can be different across evolution. Some of them will be described below (see paragraph 4.1. What drives the specificity and durability of innate immune memory?) and entail both the specificity of the memory responses and their heritability. Among the various mechanisms underlying the establishment of innate memory, it is worth mentioning endoreplication, i.e., genome duplication in cells that do not undergo mitosis, probably as wider source of DNA when RNA and protein synthesis are required at very high levels (48). In mosquito and human cells, DNA duplication was observed as being essential for the upregulation of genes associated to trained immunity (such as phenoloxidase in mosquito cells and TNFA in human monocytes) upon priming with zymosan or P. berghei (mosquito) or β-glucan (human), while modulation of genes associated to tolerance, in monocytes primed with LPS, did not depend on DNA synthesis (49). This underlines the notion that innate memory is based on diverse mechanisms, which define the type, extent and duration of the memory reaction. These mechanisms can be either homologous across evolution, as in the above case of endoreplication underlying a potentiated memory response, or widely different, both across evolution and between different priming agents in the same organism.

In mammals, sophisticated specific immunological tools, such as antibodies and T and B cell receptors, have developed thanks to gene recombination and subsequent generation of diversity. The characteristics of adaptive or acquired immunity, as it is called, are the high specificity that allows precise recognition of very short epitopes, the capacity of developing a faster and more potent memory response, based on the persistence in a quiescent state of some T and B lymphocytes after a first specific immune response, and the long duration of the memory, which is stable as it based on gene recombination and not only on epigenetic changes.

Typically, it is thought that innate immunity has an initial protective role, having the advantage of being readily active (no need for gene rearrangement). Innate immunity can take care of the vast majority of invading microorganisms, as well as endogenous senescent, dying or transformed cells and misfolded/aggregated proteins (e.g., amyloid beta and tau protein aggregates in the brain), using a wide array of cellular effectors (in particular phagocytes such as macrophages and secretory cells such as mast cells) and soluble toxic mediators (in particular complement, reactive oxygen and nitrogen species and antimicrobial peptides). Adaptive immunity is involved at a later stage, only for tackling pathogens that could not be eliminated by innate immune mechanisms.

It is however clear that the two systems are not really separate in time and functions, and that they cross-talk constantly for coordinating the global protective response. Their functions are similar in that innate immune cells can also “learn” and react in a more effective fashion after having experienced an initial reaction, although the memory mechanisms of innate immunity are different from those of adaptive memory and their range and specificity of recognition is also different (broader in innate memory, highly specific in adaptive memory) and complementary. An innate immune memory, not based on gene recombination, is also present in mammals and has been reported, besides the abundant evidence in mice and humans, in other species including non-human primates, dogs and cattle (50–52), as well as in other vertebrates such as fish (53). The first well-documented descriptions of innate memory in mammals date back to the ‘40ies - ‘60ies of last century with the description of non-specific macrophage-dependent acquired resistance to infections in mice, in particular the seminal works of Paul Beeson (54), Rene Dubos (55) and the Youmans (56). In vivo and in vitro “priming” of mouse macrophages for a better secondary response was abundantly reported from the ‘70ies by several groups, including our own (e.g., 57–61), and was recently revamped and well developed by Netea group and dubbed “trained immunity” (13, 62, 63).

Innate immune memory is mainly based on epigenetic and metabolic changes (13, 14, 64, 65). Since epigenetic modifications can be transmitted to daughter cells and induced in hematopoietic progenitors in the bone marrow microenvironment, it is to be expected that innate memory can persist in tissues beyond the lifespan of the “primed” innate cells. However, epigenetic changes can also be rapidly eliminated by multiple mechanisms or replaced by different modifications upon repeated challenges. Thus, the memory of innate cells appears to be highly plastic, as it can constantly re-program the cellular functions in response to different environmental and microenvironmental cues, so as to rapidly adapt to the external and endogenous changes.

Several other mechanisms, in addition to epigenetic and metabolic reprogramming, are involved in the establishment of innate immune memory. For instance, RNA interference (see below) is a protective mechanism that appears to be maintained across evolution. A mechanism that affords short-term memory of previous inflammatory stimulation is the modulation of the NFκB dynamics (and consequent transcriptional response) through engagement of negative feedback modules, thereby inducing an adaptive response to new stimuli (66).

Mammalian innate immune memory also appears to have some degree of specificity. The recognition tools of innate immunity, for instance TLR and scavenger receptors, are not highly specific as in the case of antibodies, but are nevertheless able to recognize different molecular patterns that characterize several classes of microorganisms. Different innate receptors can concomitantly bind different molecules on the same microorganism, thereby establishing a strong multiple interaction that blocks the target and initiates its destruction. Thus, specificity of innate immune cells is in principle determined by the suite of receptors engaged in the interaction. The lack of recombination and generation of diversity, typical of adaptive immunity, is replaced by the abundance of pre-formed germline receptors and by multiple concomitant recognition by such receptors, as in the case of innate immunity in invertebrates (6, 7, 67). Also, NFκB-based memory responses are different depending on the identity and dose of the priming stimulus (66). The features of the initial recognition (e.g., the suite of engaged receptors and triggered signaling pathways) are likely responsible for the types of epigenetic and metabolic changes that will determine the features of a secondary memory response. Thus, priming with LPS typically induces a “tolerance” type of memory response (less production of inflammatory factors), while priming with β-glucan or BCG induces a potentiated memory response (63, 68), probably exploiting the persistence of the priming agents (BCG can survive 1–2 months within macrophages, and β-glucan is hardly degraded after being phagocytosed by macrophages) (2). Notably, as also observed in invertebrates, the changes induced by priming agents do not determine a particular type of non-specific secondary response (tolerance or immunoenhancement), but rather set the basis for a more specific response that depends on the challenging agent. An example is provided in Figure 1 , showing the memory response of human monocytes primed in culture with different agents (LPS, the zymosan yeast particles and β-glucan) and later challenged with LPS or zymosan. The memory response, measured as production of three different inflammation-related factors, shows that priming with LPS or zymosan (made of insoluble β-glucan) induced an identical secondary response (decreased TNF-α, increased IL-8, unchanged IL-1β) independently of challenge, whereas priming with β-glucan induced a memory response that depends on the challenge (tendency to increased TNF-α and IL-8 and unchanged IL-1β in response to an LPS challenge; tendency to increased TNF-α and IL-8 and a substantial increase of IL-1β in response to a zymosan challenge) ( Figures 1A-C ). Thus, a challenge with zymosan specifically induces an enhanced response in terms of IL-1β production, whereas the potentiated response in terms of TNF-α and IL-8 production seems to be less evident and non-specific. It is notable that the same substance (β-glucan) can induce different types of memory depending on its form (insoluble zymosan particles vs. soluble molecule), suggesting that the induction of memory is modulated by the size/solubility of the priming agent. It is also important to note that the same combination of priming and challenge can generate a different memory response in monocyte-derived macrophages. Priming with LPS or zymosan induced a tolerance response to LPS or zymosan in terms of TNF-α production, but no significant change in IL-8 and IL-1β ( Figures 1D-F ). Eventually, priming with β-glucan induced no change in any of the three cytokines in response to either LPS or zymosan ( Figures 1D-F ). Thus, each particular combination of priming and challenging stimuli can induce a specific memory response profile, which can also differ between different innate cell types. Thus, the innate memory response is in its way antigen-specific, as it is based on the cell-specific and priming/challenge-specific combined increased or decreased production of a range of inflammatory/immunostimulatory and anti-inflammatory/immunomodulatory factors that is expected to optimally support a defensive reaction to the new challenge.

Another notable feature of innate memory specificity in mammalians relates to organ tropism. There is evidence both in mice and humans that exposure to lung infectious agents can induce an innate memory that will protect against the same or other lung-specific infections or diseases (e.g., asthma), but not, or less potently, against infections/allergies typical of other body compartments (69–72). Although this is not always the case, and systemic immune memory can be generated after priming in different anatomical compartments, this suggests that the organ-specific microenvironment can play a role in the establishment and maintenance of topically protective innate memory.

The original contributions presented in the study are included in the article/ Supplementary Material . Further inquiries can be directed to the corresponding author.

The studies involving humans were approved by Regional Ethics Committee for Clinical Experimentation of the Tuscany Region (Ethics Committee Register n.14,914 of 16 May 2019). The studies were conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study.

DB: Conceptualization, Data curation, Funding acquisition, Writing – original draft, Writing – review & editing. ET: Investigation, Methodology, Writing – review & editing. PI: Conceptualization, Data curation, Investigation, Methodology, Writing – review & editing.