During the last several years, our knowledge on the interleukin (IL)-1 family molecules, as central mediators of innate immunity/inflammation and as “guilty” molecules of the development of autoinflammatory, autoimmune, infectious and degenerative diseases, has increased (Sims and Smith, 2010; Dinarello, 2011a,b, 2013; Dinarello and van der Meer, 2013; Garlanda et al., 2013a) The IL-1 family encompasses 11 cytokines/ligands and 10 related receptors (Dinarello et al., 2010; Boraschi and Tagliabue, 2013). Among the IL-1 family ligands, IL-1β is produced by mononuclear phagocytes in response to infectious or other stressful events, and initiates a potent defensive inflammatory response, while the structurally similar IL-1α is released only upon cell death and functions as an alarmin (Dinarello, 2011a; Rider et al., 2013). The IL-1-induced inflammation is regulated by a complex interaction of receptors and soluble inhibitors, whose concerted action determines the timing of activity and its shut-off. Both IL-1α and IL-1β bind to IL-1R1 and form an activating complex with the signaling chain IL-1R3. The receptor antagonist IL-1Ra binds to IL-1R1 receptor with high affinity, thereby competing with IL-1α and IL-1β, and does not recruit IL-1R3 (thus the complex is inactive). The other IL-1-binding receptor, the decoy receptor IL-1R2, can bind IL-1β and less efficiently IL-1α and IL-1Ra, and can recruit IL-1R3. However, the complex is inactive due to the lack of signal-initiating sequences in the intracellular domain of IL-1R2. The soluble forms of these receptors (sIL-1R1, sIL-1R2, sIL-1R3) have an inhibitory function by acting as ligand traps, and ensure a balance between amplification/activation of defensive responses and uncontrolled inflammation (Boraschi and Tagliabue, 2013; Garlanda et al., 2013b).

IL-1β is primarily produced by hematopoietic cells in response to various microbial stimuli, activated complement components, other inflammatory cytokines (e.g., TNF-α) and IL-1 itself. IL-1β is synthesized as a long inactive pro-form, which needs cleavage within the context of the inflammasome for being activated and then secreted (by non-conventional mechanisms) (Burns et al., 2003; Martinon et al., 2009; Monteleone et al., 2015). Activation of the inflammasome and of the IL-1 cleaving enzyme caspase-1 is therefore an additional mechanism controlling IL-1β-induced inflammation.

IL-1β production may vary, in innate immune cells, depending on the activation status of such cells. The concept of innate immune memory, i.e., the variation of innate reactivity in cells previously exposed to various stimuli, is a concept well known in invertebrates and also in vertebrates, which has been recently re-confirmed in higher vertebrates and humans (Kleinnijenhuis et al., 2012; Netea et al., 2016). A re-programming of innate immune cells can lead to decreased (tolerance) or enhanced (training) reactivity against reinfection by the same or different pathogens. Tolerance aims to avoid extensive tissue damage, whereas training aims to improve tissue surveillance, necessary to protect weakened tissues (Ifrim et al., 2014; Töpfer et al., 2015). It has been known for several decades that priming of mononuclear phagocytes with lipopolysaccharide (LPS) inhibits cellular functions in a process called LPS-induced tolerance (Dobrovolskaia and Vogel, 2002; Fan and Cook, 2004), whereas only recently it has been shown that priming with Candida albicans or the fungal cell wall component β-glucan can induce enhanced responses (Quintin et al., 2012). The effect of different types of bacteria (Gram-negative vs. Gram-positive) on the development of innate immune memory still remains unclear. While Gram-negative bacterial components such as LPS can lead to innate tolerance and a decreased response, exposure to Gram-positive Bacillus Calmette–Guérin leads to trained immunity and a more effective host immune response, accompanied by a reduced mortality to non-related infections (Quintin et al., 2014; Blok et al., 2015).

A series of recent in vitro and in vivo experiments has shown that pathogen-associated molecular patterns and a number of danger-associated molecular patterns induce innate immune memory (Netea et al., 2011; Crisan et al., 2016). Molecular and cellular mechanisms involved in this phenomenon are still not fully understood. It is believed that the main mechanism underlying enhanced responses involves epigenetic reprogramming, which in turn may entail altered pattern recognition receptors’ expression, metabolic reprogramming, or/and altered cytokines release (Saeed et al., 2014; Bekkering et al., 2016b). Long-term epigenetic changes in monocytes and macrophages apparently involve both histone methylation and acetylation, as for instance H3K4 monomethylation and H3K27 acetylation induced by LPS (Ostuni et al., 2013; Saeed et al., 2014), and histone H3K4 trimethylation and H3K27 acetylation caused by β-glucan (Quintin et al., 2012; Saeed et al., 2014). Recently, it has been hypothesized that innate memory could also involve the modulation of expression of “latent and de novo” enhancers, microRNAs and/or long non-coding RNAs (Netea et al., 2016).

Unraveling the mechanisms at the basis of innate memory could lead to a better understanding of innate host defense and development of new immunization strategies and immunotherapies (Töpfer et al., 2015).

The aim of the present study was to investigate how innate memory can change the inflammatory reactivity of human monocytes and macrophages. Our study particularly focuses on changes in the levels of active available IL-1β produced by cells exposed and re-stimulated with Gram-positive and Gram-negative bacteria and bacterial components. This could set the stage for understanding how innate memory could be exploited for regulating IL-1 family ligands and receptors for innovative therapies of IL-1-mediated diseases, as well as new immunization strategies.

Regulation of the activity of IL-1β is of paramount importance for obtaining optimal defensive reactions without causing excessive tissue damage. Indeed, the mechanisms of IL-1 regulation are multiple and at all levels, from regulation of its maturation/secretion to regulation of its activity both at the receptor binding level and at the level of post-receptor signaling. This kind of regulation reflects the powerful biological activity of IL-1β that, if not properly controlled, can be at the basis of numerous inflammatory and degenerative diseases (Dinarello, 2011a,b).

In this study, we have investigated a novel mechanism of IL-1β regulation, through the generation of innate memory.

Several studies have recently revived the old concept of innate immune memory, i.e., the ability of innate immune cells to “remember” previous encounters with microorganisms or foreign agents by changing their reactivity to a subsequent challenge with the same or with a different agent (Netea et al., 2011; Saeed et al., 2014; Gardiner and Mills, 2016). The concept, well known in invertebrates and also in mice, applies also to human innate immune cells in vivo and in vitro (Quintin et al., 2012; Arts et al., 2015; Bekkering et al., 2016a; van der Valk et al., 2016). In our study, we have used an in vitro system, based on human primary monocytes and tissue-like/deactivated monocyte-derived macrophages from the same donors, to examine the role of innate memory in determining the production of active IL-1β. We have used as stimuli two bacteria, the Gram-negative E. coli and the Gram-positive L. acidophilus, and a molecule derived from E. coli, i.e., its cell wall LPS. Data presented here were obtained by using the same stimulus for both priming and challenge of cells in culture. However, preliminary data of cross-stimulation confirm the notion that innate memory is not stimulus-specific (Figure 3).

The first observation we made is that monocytes are more reactive than macrophages to microbial stimuli in terms of production of inflammatory cytokines and chemokines (Figure 1). Although this was expected, it is noteworthy that while the difference in the production of inflammatory cytokines (IL-1β, TNF-α) is significant (up to almost 100×), there is much less difference in the production of chemokines (CXCL8, CCL2). Weaker response of macrophages in terms of inflammatory effector cytokines but almost normal response in terms of alarm signals (such as chemokines) is in line with the sentinel role of tissue-resident macrophages, which should be active in producing chemokines such as CXCL8 and CCL2, aiming at recruiting effector cells (e.g., neutrophils and blood circulating monocytes), but not directly involved in the very early effector phase of inflammation, in order to avoid uncontrolled tissue damage (Davies et al., 2013).

The innate immune memory appears as a decreased (tolerance) or increased response (trained innate immunity) to a second immune challenge. Both tolerance and trained innate immunity seem to be the result of long-term epigenetic changes in monocytes and macrophages, and the ability of these cells to “remember” is actually dependent on epigenetic changes. All the epigenetic changes and molecular mechanisms at the basis of memory eventually result in lower or higher transcription levels in several genes, such as pathogen recognition receptors, signaling molecules, and cytokines, in a short window of time. For example, priming of human monocytes with IFN-γ or IL-10, or the interaction with microbial components, alters the receptor repertoire expressed by monocytes/macrophages, e.g., changes in TLR4 and MD2, and in Dectin-1 and MARCO in macrophages (Bosisio et al., 2002; Willment et al., 2003), or in signaling molecules such as MyD88 (Bosisio et al., 2002). These changes in the number of receptors and the regulation of signaling drive a different production of effector molecules, such as inflammatory and anti-inflammatory cytokines. Although in this study we did not investigate the molecular mechanisms underlying the difference of responsiveness between monocytes and macrophages in terms of cytokine production, it is possible to speculate that the two cell types may undergo different epigenetic changes. In fact, although macrophages in this study are derived from monocytes, the in vitro differentiation protocol with deactivating cytokines and growth factors yields a cell population that differs from monocytes also in terms of the level of innate receptors (Ifrim et al., 2014). It is plausible that chromatin modifications occur during the differentiation process, and, when macrophages are primed and challenged, this is translated in a transcription programs different from that of monocytes.

In our in vitro model of innate memory, we observed two important facts. First, that also in the case of memory there is little effect on chemokine production, suggesting that the ability of cells to send alarm signals should not be hampered. A second observation is that memory is not induction of tolerance or induction of enhanced responses. In fact, the same stimulus can induce decreased or increased responses in primed cells depending on the endpoint measured (production of one or another of the various inflammation-related factors) and on the cell type examined (monocytes vs. macrophages). Thus, innate memory is a complete re-programming of the reactivity of cells rather than a decreased or enhanced responsiveness.

We focused on how innate memory could affect the production of active IL-1β. For doing this we have separately measured the production of IL-1β and of a number of soluble factors, concomitantly produced by monocytes and macrophages, that could affect its biological activity. The inflammatory activity of IL-1β is similar to that of IL-1α, which, however, is a cytokine mostly active intracellularly or as a cell-associated factor in the cell–cell communication (Kim et al., 2013; Di Paolo and Shayakhmetov, 2016; Bertheloot and Latz, 2017). Indeed, we found that IL-1α was practically absent in the supernatants of monocyte and macrophage cultures stimulated with the various stimuli (data not shown). Thus, the IL-1-like activity in these supernatants can be exclusively attributed to IL-1β. It is known that the soluble form of IL-1R2 is a major inhibitor of IL-1β, as it can capture it in solution with high affinity (2.7 nM), while unable to bind with sufficient affinity either IL-1α (1.6 μM) or IL-1Ra (25 μM) (Symons et al., 1995). Thus, sIL-1R2 can be considered as a specific inhibitor of IL-1β. Monocytes and macrophages also produce significant levels of the soluble forms of the other two IL-1 receptor chains, i.e., sIL-1R1 and sIL-1R3. The IL-1-binding sIL-1R1 possibly binds IL-1Ra with higher affinity than IL-1α and IL-1β (Arend et al., 1994), while sIL-1R3 might form complexes with sIL-1R2 of higher affinity for IL-1α and IL-1β but not IL-1Ra (Smith et al., 2003). In these circumstances, since the levels of both sIL-1R1 and sIL-1R3 were constant in monocyte and macrophage supernatants and did not depend on stimulation or memory (Table 1 and data not shown), we decided to exclude them from the calculation of IL-1β activity. Thus, we have limited the calculation to two elements, IL-1β and sIL-1R2, binding to each other at a 1:1 stoichiometric ratio and with a known affinity (2.7 nM). It was therefore possible to apply the law of mass action to such interaction and to calculate the amount of free IL-1β, i.e., the amount of cytokine not blocked by sIL-1R2 and therefore available for activating target cells by binding to membrane receptors. However, since the presence of IL-1Ra is expected to compete with free IL-1β for receptor binding, we have also taken into account its presence, in order to have a more thorough evaluation of biologically active IL-1β. IL-1Ra is abundantly produced by monocytes (0.9–3.0 ng/ml) and more abundantly by macrophages (9–150 ng/ml), is upregulated upon stimulation (2–7×; data not shown) and it varies depending on memory (Figure 4). Thus, its presence is relevant for determining the final capacity of IL-1β to exert its inflammatory effects. In the absence of reliable data for accurately calculating possible competition between the two ligands in a complex system such as the membrane of a responding cells (in which activation is regulated not only by the quantitative presence of the receptor chains but also by the changes in affinity due to the presence of accessory chains and inhibitory receptors), we have indicatively expressed the levels of active IL-1β as the ratio between free IL-1β and IL-1Ra. We have observed several very interesting phenomena. The first observation is, as expected, that each individual subject responds differently, both in quantitative terms (the amount of cytokine produced) and also in terms of type of response (enhanced reaction vs. decreased response). This behavior most likely depends on the past history of exposure of the donor, i.e., age, vaccinations, diseases, etc. Another important observation is the different behavior of macrophages (the tissue-resident mononuclear phagocytes) as opposed to monocytes (the blood-borne inflammatory cells). As an example, monocytes of donor 1 reacted to challenge with E. coli with a sharp increase in active IL-1β, whereas macrophages of the same donor significantly down-regulated active IL-1β production upon challenge with E. coli. This underlines the different role of the two cell types in inflammatory reactions, with monocytes being the inflammatory effector cells that must be ready and reactive, whereas tissue macrophages must avoid excessive reaction that would cause unwanted tissue damage. Another observation that warrants attention is the different response raised by whole E. coli bacteria as opposed to the bioactive E. coli LPS molecule. For instance, in the case of donor 2, active IL-1β produced by unprimed monocytes in response to LPS is lower than that induced by E. coli and is not changed by priming, whereas response to E. coli is significantly reduced by priming. Macrophages of the same donor are responsive to LPS but practically unresponsive to E. coli, and their response is reduced by priming with LPS but not by priming with E. coli. Also, it is notable that macrophages seem to respond either to whole E. coli or to LPS, but not to both (see donors 1 and 3: no response to LPS vs. significant response to E. coli; and donor 2: good response to LPS and undetectable response to E. coli).

Our conclusions are based on a preliminary assessment on three donors, and the very high inter-individual variability of the findings may lead to over-interpretation of the data. On the other hand, since the reactivity of the immune system is tailored on previous experiences, it is in a way expected that each subject could respond differently, based on his/her individual immunological history. Thus, increasing the number of subjects may not overcome the issue of inter-individual variability. What we would like to underline with these data is the importance of an individual evaluation of the innate memory status, in view of future approaches of precision medicine that could help us improving/optimizing immunotherapeutic strategies. In cases as that of innate memory, it seems clear that no general conclusions can be drawn, and in these circumstances a case-by-case assessment becomes crucial.

Overall, it could be concluded that innate memory limits the amount of active IL-1β produced by tissue macrophages in response to a challenge, in line with the hypothesis that these cells are sentinels and not effector cells, and should avoid damaging the tissue by initiating a potentially destructive inflammatory response. On the other hand, memory tends to enhance the reaction of monocytes, which are the effector cells in inflammatory reactions, thereby making them more efficacious in combating the potential danger. Thus, the induction of innate memory could help increasing the host resistance to infections without causing excessive local tissue damage. This could be among the reasons for the efficacy of adjuvants in enhancing protective immunity, in addition to facilitating the establishment of adaptive memory. It is exciting the hypothesis that manipulation of innate memory may also become an important therapeutic strategy in chronic inflammatory and degenerative diseases.

MM performed the experiments and wrote the paper; ET performed cross-stimulation experiments; PI and DB designed the study and wrote the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.