BCG can be used as a training agent via modulating intracellular signaling pathways by binding to both toll like receptor 2 (TLR2) and TLR4 to induce trained immunity (20). Subsequently, training monocytes with BCG activates the mammalian target of rapamycin (Akt/mTOR) pathway, leading to upregulation of glycolysis, oxidative phosphorylation (OXPHOS), and glutamine metabolism (21). In addition to the Akt/mTOR pathway, the nucleotide-binding oligomerization domain 2 (NOD2) pathway [activated via TLR2/4, which can induce IL-6, IL-1β, and tumor necrosis factor (TNF) (22, 23)], was found to be necessary for BCG-training of monocytes: blocking NOD2 by mutation or chemical inhibition resulted in BCG-trained monocytes that produced significantly lower amounts of TNF upon restimulation (24).

A study of severe combined immunodeficiency (SCID) mice treated with BCG demonstrated enhanced release of cytokines IL-1β and TNF in monocytes in a NOD2-dependent manner upon infection with Candida albicans, which was found to be epigenetically controlled by histone 3 lysine 4 trimethylation (H3K4me3), an open chromatin histone marker (24). Similarly, the Akt pathway was involved in BCG-trained neutrophils, inducing increased H3K4me3 patterns, specifically near promoter regions of genes encoding pro-inflammatory cytokines (CXCL8 and IL-1β) and genes involved in glycolysis (mTOR and PFKB); this lasted at least three months post-vaccination (25). Epigenetic modifications at genes involved in signaling pathways such as AKT1, MAPKs, and PI3K (26) are responsible for inducing trained immunity and have been discovered within the epigenome of specific effector cells such as monocytes (27). This suggests that global epigenetic modifications also directly influence the “trained” functions of specific innate immune cells. Modulations in gene expression result in the increased production of cytokines involved in trained innate immunity. For example, IL-1β appears to be a critical cytokine induced by BCG-training: in BCG-trained monocytes, IL-1β levels increased after restimulation conferring non-specific protection against yellow fever viremia (26). The antimicrobial cytokine IL-32, which is known to be involved in viral defense and capable of modulating cellular metabolism (28), has also been demonstrated to play a role in BCG trained immunity, presumably by activating the NOD2 signaling pathway (29) ( Figure 1 ).

Taken together, signaling pathways and epigenetic modifications induced by BCG-training are responsible for the enhanced innate immune response upon restimulation. Many of these signaling pathways are activated in different cell types within the innate immune system, suggesting that training with BCG may utilize conserved pathways. Since BCG vaccination has been demonstrated to possess therapeutic effects against respiratory infections, it has been hypothesized that BCG may offer heterologous immunity against SARS-CoV-2 infection (17, 30, 31). Therefore, ongoing clinical trials are investigating the relationship between BCG vaccination and protection against COVID-19. Currently, the data suggests that BCG-trained immune responses are present 3 months post BCG-treatment (24); thus, further research should be conducted to evaluate the cellular response of BCG-trained cells for longer times to determine the maximal duration of memory response and protection.

β-glucan is an immunomodulator that functions as a pathogen associated molecular pattern (PAMP) recognized by pattern recognition receptors (PRRs) expressed on innate immune cells. The immunomodulatory activity of β-glucan not only depends on its physical properties (for example, particle size, molecular weight, conformation, branching frequency, or solubility) but also on the cell type it acts upon, suggesting a highly diverse and targetable approach for β-glucan training (31). β-glucan signaling occurs in monocytes, macrophages, and DCs by the dectin-1 receptor, which activates an intracellular signaling cascade that drives the trained immune response (32–34). Upon β-glucan binding to dectin-1, the downstream Akt/mTOR/HIF1α pathway is activated, leading to subsequent metabolic modifications, including the downregulation of OXPHOS and upregulation of glycolysis (35), resulting in the increased production of tricarboxylic acid (TCA) cycle by-products, including succinate (36), fumarate (37), and mevalonate (a precursor for cholesterol synthesis) (38). In addition to binding to the dectin-1 receptor, β-glucan has been suggested to utilize other receptors to induce trained immunity (39).

Intracellular modifications of the TCA cycle driven through the Akt/mTOR/HIF1α pathway were found to play a key role in β-glucan trained immunity. Training with β-glucan was associated with increased levels of succinate, which has been suggested to have potential therapeutic benefits against sepsis-induced immunoparalysis (36). Fumarate, a metabolite involved in the TCA cycle, was found to increase cytokine production in a dose-dependent manner by inhibiting KDM5 histone demethylases, thereby leading to epigenetic changes (trimethylation, H3K4me3) at promoter regions of pro-inflammatory cytokine genes (37). Interestingly, KDM5B also regulates angiotensin-converting enzyme 2 (ACE2), the receptor involved in SARS-CoV-2 viral entry (40), providing a potential therapeutic target. Fumarate also stabilizes HIF1α by inhibiting hydroxylation, resulting in the induction of 2,688 dynamic H3K27ac changes, as revealed by chromatin immunoprecipitation followed by sequencing (ChIP-seq) (37). Monocyte training with mevalonate, indirectly synthesized from acetyl CoA, increased cytokine production upon restimulation. Importantly, inhibition of mTOR signaling or glycolysis impedes mevalonate training (38). Thus, β-glucan trained immunity is induced through distinct signaling pathways which drive metabolic and subsequent epigenetic modifications to maintain a sustained, non-specific response to invading pathogens ( Figure 1 ).

Epigenetic changes induced by β-glucan in hematopoietic stem and progenitor cells (HSPCs) enable short-lived blood circulating monocytes to respond to secondary challenge, which leads to an increased level of IL-1β in trained monocytes, further inducing the production of IL-32 (41). Similarly, epigenetic reprograming of HSPCs by β-glucan drives trained immunity in an IL-1 dependent manner (42). Thus, β-glucan has a strong potential to be used as a training agent and already has prior applications as an oral supplement, which have been discussed in preventing SARS-CoV-2 infection (43). For example, since IL-1 is involved in COVID-19 dysregulated immune responses, recent studies have determined that blocking its receptor (IL-1R) can be used to treat severe COVID-19 infection (44). Since IL-1β production is greatly increased during COVID-19 (45), β-glucan trained immunity may enable the innate immune system to mount a stronger early immune response to prevent dysregulated cytokine production. This training of innate immune cells to increase production of key cytokines may ultimately improve the immune response upon SARS-CoV-2 infection.

LPS, a microbial component comprising gram-negative bacterial membranes, binds to TLR4 and has primarily been used to restimulate cells after BCG or β-glucan training, but can also elicit trained immunity under certain conditions (46). Innate immune cells can be primed, resulting in transcriptional changes that provide an increased basal innate immune response (12). Prolonged priming with LPS results in reduced cytokine production and immunosuppression, inducing immune tolerance (47). These effects are due to reduced expression of TLR4, reduced TLR4/MyD88 interaction, and reduced activity of p38 (47). LPS immune tolerance was found to block the accumulation of histone marks at promotors or enhancers (H3K27ac and H3K4me1) of genes that are responsible for lipid metabolism and phagocytic pathways (48). LPS triggers transient silencing of inflammatory genes (associated with methylation at H3K9 and H3K27) as well as inducing antimicrobial effector genes (49). Training with LPS has been demonstrated to reduce inflammation and fibrosis in systemic sclerosis models by downregulating production of proinflammatory cytokines (IL-6, TNF, IL-1β) and instead upregulating production of the anti-inflammatory cytokine IL-10 (50, 51).

Conversely, lower doses of LPS treatment result in an increase of TNF and IL-6 upon restimulation (52), suggesting a role for LPS in trained immunity. LPS can induce trained immunity by activating innate immune cells like macrophages through chromatin and transcriptional modifications via signaling through the MAPK/p38 pathway ( Figure 1 ) (11, 52). These include epigenetic modifications in myeloid cells, where treatment with LPS is associated with chromatin remodeling, such as an increased number of activating marks (53). Evidence suggests that these changes may be long lasting, since LPS stimulation of macrophages induced persistent methylation at H3K4 that corresponded to an enhanced response upon secondary stimulation (54). Additionally, LPS can induce trained immunity through regulation of transcription factors, like ATF7. LPS-treated macrophages were shown to induce phosphorylation and subsequent release of ATF7 via the p38 pathway, resulting in a decrease of repressive H3K9me2, which persists for long periods of time to increase resistance to subsequent reinfection (55). LPS has also been identified to induce epigenetic changes within hematopoietic stem cells (HSCs) that are dependent on the transcription factor C/EBPβ, which likely facilitates the long-lasting transcriptional changes shown to be induced by LPS (19). Similar to β-glucan, LPS treatment has been demonstrated to downregulate OXPHOS and upregulate glycolysis, resulting in increased levels of succinate (56), suggesting that LPS may induce trained immunity through multiple signaling pathways.

Finally, LPS has been demonstrated to indirectly decrease ACE2 expression via NF-κB-dependent upregulation of miR-200c-3p during H5N1 virus infection (57), suggesting that LPS may have additional preventative and therapeutic roles in COVID-19. Taken together, LPS training may provide beneficial effects to impair the ability of viral infectiousness and reduce severe inflammation in COVID-19.

In addition to BCG, β-glucan, and LPS, alternative immune training agents have been investigated, including aldosterone and oxidized low-density lipoprotein (oxLDL). Aldosterone-trained monocytes were characterized by enriched H3K4me3 at the promoter regions of genes involved in lipid synthesis (58). Similar to BCG treatment, oxLDL training upregulates both OXPHOS and glycolysis (59), which similarly results in increased TCA metabolites (60). Increased levels of proinflammatory cytokines IL-6 and TNF were mediated by the mTOR/HIF-1α pathway (61). These data suggest that oxLDL shares similar yet distinct pathways with BCG and β-glucan training, providing the opportunity to selectively train innate immunity by modifying metabolic pathways.

In contrast, several studies have highlighted that some reagents function as training antagonists, which can function to inhibit cell signaling or alter epigenetic modifications required to induce trained immunity. For example, BCG-incubated monocytes co-incubated with ARTA (a vitamin A metabolite) displayed reduced cytokine production, suggesting the role of ARTA as a trained immunity antagonist. This unique signature was determined to be caused by both increased expression of SUV39H2, a histone methyltransferase that induces the inhibitory histone 3 lysine 9 trimethylation (H3K9me3), and by downregulation of the stimulatory H3K4me3 (62). In addition to ARTA, some cytokines may also function as trained immunity antagonists. The anti-inflammatory cytokines IL-37 and IL-38 as well as the drug Hydroxychloroquine have demonstrated inhibitory effects against trained immunity. Both IL-37 and IL-38 inhibit β-glucan-mediated trained immunity by blocking mTOR signaling (63). IL-37 inhibits β-glucan trained immunity by reversing immunometabolic and histone modifications, thus suppressing production of pro-inflammatory cytokines (64). Likewise, hydroxychloroquine acts through the Akt/mTOR pathway to inhibit β-glucan trained immunity by preventing the production of TNF, IL-6, and IL-1β cytokines via blocking epigenetic modifications (65). Taken together, these trained immunity antagonists may be used with training reagents simultaneously to regulate the level of the immune response.

Monocytes are critical innate immune cells that have a relatively short lifespan, circulating in the blood stream for 3 to 5 days (76). Despite their short lifespans, monocytes have been effective in presenting trained immune responses, which are driven by epigenetic histone modifications and have been shown to last at least 3 months (24), suggesting that monocyte reprograming may occur within the bone marrow and progenitor cells (11). Recent studies have classified trained immune responses as belonging to central trained immunity (occurring in bone marrow progenitor cells) and peripheral trained immunity (in blood monocytes and tissue macrophages). Monocyte reprogramming occurs at both levels, first by reprograming HSCs in the bone marrow followed by migration and potential differentiation into macrophages within the blood (74).

Epigenetic modifications after monocyte training drive subsequent changes in metabolic signatures. β-glucan trained monocytes exhibit increased glucose consumption and a shift from OXPHOS towards aerobic glycolysis (35), similar to what is observed in cancer metabolism (77). Epigenetic changes may be more dynamic than a few chromatin modifications, as training monocytes with β-glucan caused histone acetylation changes which resulted in approximately 8,000 dynamic regions (51). Taken together, these studies demonstrate that epigenetic modifications are responsible for sustained training effects in monocytes.

Monocytes have been shown to be involved in the innate immune response to COVID-19 (78). Several pro-inflammatory proteins, such as CXCL-10, IL-6, IL-10, IL-12, and TNF, have been identified to be upregulated in severe COVID-19 patients. Upregulation of these chemokine and cytokines were related to increased leptin, a hormone that signals through the NF-κB and STAT3 pathways (79). Similar to the effects of trained immunity, epigenetic changes such as open chromatin marks were present in monocytes from acute-COVID-19 patients and persisted six months after recovery from COVID-19 (80). Peripheral monocytes from COVID-19 patients exhibited mitochondrial dysfunction as evidenced by reduced glycolysis (81). Since trained immunity in monocytes is characterized by an upregulation of glycolysis (35), in vivo training of monocytes may help counteract this reduced glycolysis in COVID-19 patients. Identifying similar monocyte signatures during SARS-CoV-2 infection and innate immune training may help determine a role for trained immunity to prevent or treat COVID-19.

Although NK cells are part of the innate immune system, they also share certain memory characteristics of adaptive immune cells (11, 82). NK cells can respond to host infections by recognizing cytokines IL-12 and IL-18 and producing interferon-gamma (IFN-γ) and TNF, as well as excreting perforin and granzymes which penetrate the membrane of an infected cell and induce apoptosis (83). Most innate immune cells are typically considered to be short lived; however, NK cells can live for longer than 6 months in both lymphoid and non-lymphoid organs (84). Similar to cytotoxic T cells, it has been shown that NK cells specific to a cytomegalovirus (CMV) virus receptor have a lifespan of several months (85). Memory NK cells are distinguished by their abilities of enhanced mitochondrial fitness, regulation, and OXPHOS capacity (86).

Although NK cells trained by BCG have been shown to upregulate the production of pro-inflammatory cytokines (IL-1β, IL-6, TNF) (16), the specific mechanism by which trained immunity is induced in NK cells is still poorly understood. The general mechanisms of NK memory and proliferation have been studied in murine models. Not surprisingly, many NK memory-like functions are driven by epigenetic modifications. In CMV infected mice, NK cells were able to mount adaptive immune functions which were subsequently identified to be a result of DNA hypermethylation patterns, similar to cytotoxic T cells (87). In vitro stimulation of NK cells with cytokines IL-12 and IL-18 followed by adoptive transfer into immune deficient mice led to memory-like NK responses that lasted for three months (88). Similar to trained monocytes which are characterized by metabolic reprogramming, recent studies of NK cell metabolism have shown that key metabolic energy pathways such as OXPHOS and glycolysis drive NK anti-viral responses. Moreover, NK cells shift from cytotoxic proliferation to a quiescent state upon resolution of inflammation, indicating that NK cells are highly dynamic which is beneficial for the trained immunity process (86). Metabolic analysis demonstrated that after exposure to a high concentration of IL-15, NK cells became activated via the mTOR pathway; inhibition of this pathway rendered NK cells incapable of proliferating and mounting cytotoxic responses (89). Finally, NK cells may play an indirect role in trained immunity by modulating monocyte function. It has been determined that bone-marrow-resident NK cells promote monocyte effector functions by secreting IFN-γ (75). Alternatively, innate immune crosstalk may also occur in the opposite direction, as macrophages can activate NK cells through secretion of IL-12 and IL-18 or direct ligand interactions (90, 91) Thus, the crosstalk among innate immune cells should be further investigated in trained immunity.

Since NK cells possess memory-like characteristics, training NK cells may allow for a more robust innate response against COVID-19. In hospitalized COVID-19 patients, overall NK cell counts were low; however, the level of adaptive NK cells was increased (92). Furthermore, in severe COVID-19 patients, NK cells were functionally impaired due to elevated IFN-α and TNF signaling (93), suggesting that dysregulation of NK cells is involved in pathogenesis during SARS-CoV-2 infection. Studying the underlying mechanisms by which trained immunity may enhance NK cell function could illustrate the potential application of training NK cells as a COVID-19 therapeutic approach.