Figure 6The Role of IL-1β for Trained Immunity
(A) Genetic variations in IL-1β were assessed in DNA samples from 98 healthy volunteers from the 200FG cohort by a genome-wide Illumina Infinium SNP array. Monocytes were isolated and stimulated in vitro with culture medium (negative control) or BCG (10 μg/mL) for 24 hr. On day 6, cells were restimulated with culture medium or LPS (10 ng/mL). After 24 hr, supernatants were taken and cytokine concentrations were assessed using ELISA. Medium-restimulated cells did not produce detectible cytokine production. A SNP in IL-1β (IL1B; rs16944) was found to affect the trained immunity response induced by BCG (mean ± SEM, ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, Mann-Whitney U test).
(B) Additional genetic analysis in genes of the IL-1 pathway identify polymorphisms in the IL-1 and IL-18 receptors, as well as the inflammasome component PYCARD/ASC, that significantly modulate BCG-induced trained immunity.
(C) Schematic outline of the in vitro training experiments. Human monocytes were primed for 24 hr with culture medium or different concentrations of IL-1β. On day 6, cells were restimulated for 24 hr with LPS, and levels of IL-6 and TNFα were assessed (mean ± SEM, n = 7, ∗p < 0.05).
(D) Human monocytes were primed for 24 hr with culture medium or different concentrations of IL-1β. On day 6, cells were restimulated for 24 hr with LPS, and levels of IL-6 and TNFα were assessed (mean ± SEM, n = 7, ∗p < 0.05).
(E) Human monocytes were incubated with either culture medium, IL-1β (10 ng/mL), or BCG (5 μg/mL). At day 6, cells were restimulated with YFV (1,000 IU/mL). Concentrations of IL-8 in the supernatants were measured after 24 hr (mean ± SEM, n = 6), and mRNA expression of IL-8 was determined after 4 hr restimulation (mean ± SEM, n = 3 in duplo, ∗p < 0.05, Wilcoxon signed-rank test).
(F) Monocytes were stimulated for 24 hr with culture medium or IL-1β (100 ng/mL) in the presence or absence of MTA (1 mM). On day 6, after restimulation with LPS for 24 hr, IL-6 and TNFα were measured in the supernatants (mean ± SEM, n = 7, ∗p < 0.05, Wilcoxon signed-rank test).
(G and H) Human monocytes were stimulated for 24 hr with culture medium or IL-1β (100 ng/mL). Chromatin was fixated on day 6 and ChIP-qPCR was performed. H3K4me3 and H3K9me3 were determined at promoters of IL6, TNFA, and IL1B (mean ± SEM, n = 8, ∗p < 0.05, Wilcoxon signed-rank test).
Epigenetic Changes Mediate IL-1β-Induced Trained Immunity
To obtain further insight into the role of IL-1β in mediating trained immunity responses, we next assessed whether IL-1β itself can induce trained immunity in vitro.
Using the previously described in vitro model of trained immunity (Bekkering et al., 2016, Kleinnijenhuis et al., 2012), priming of monocytes with IL-1β (1–10 ng/mL) for 24 hr (Figure 6C) enhanced production of the pro-inflammatory cytokines IL-6 and TNFα upon LPS restimulation on day 6 (Figure 6D).
We also observed enhanced IL-8 production by monocytes that were trained with IL-1β or BCG and were restimulated with YFV.
mRNA expression analysis of IL-8 demonstrated an enhanced upregulation of IL-8 in both BCG and IL-1β trained monocytes compared to control (Figure 6E).
Restimulation with YFV on day 6 did not lead to IL-6 or TNFα production (data not shown).
Pre-treatment with the histone methyltransferase inhibitor MTA (5′- deoxy-5′-methylthio-adenosine) abrogated the IL-1β-induced trained immunity response, suggesting that the in vitro training effect of IL-1β is mediated via epigenetic reprogramming of monocytes (Figure 6F).
This was confirmed by IL-1β-mediated induction of epigenetic changes 6 days after IL-1β treatment in promoter regions of TNFA, IL6, and IL1B, with increased trimethylation of lysine 4 at histone 3 (H3K4me3; Figure 6G), a marker of increased gene transcription, and reduced trimethylation of lysine 9 at histone 3 (H3K9me3; Figure 6H), a repressor mark.
The increase of H3K4me3 and the decrease of H3K9me3 induced by IL-1β was most pronounced at the promoter of IL1B.
Recent studies demonstrated that functional adaptation of innate immune cells after an infection or vaccination, which is induced through epigenetic reprogramming, is responsible for a de facto innate immune memory, also termed trained immunity (Netea et al., 2016).
BCG vaccination has long been known to be accompanied by lower mortality in vaccinated children, an effect that cannot be fully ascribed to protection against tuberculosis, but that is probably due to protection against neonatal sepsis and respiratory tract infections.
Although it has been reported that this heterologous protection is due at least in part to induction of trained immunity (Kleinnijenhuis et al., 2012), no comprehensive assessment of the epigenetic programs induced by BCG has been done until now.
Also, evidence for protective effects of BCG-induced trained immunity on unrelated infections in humans is outstanding.
In this study, we describe the epigenetic changes induced at a genome scale by BCG in human monocytes.
We chose to study the H3K27ac histone mark because it is associated with an “active” chromatin state at both gene promoters and enhancers.
Moreover, we previously showed a strong correlation between H3K27ac and H3K4me3 changes during trained immunity, with H3K27ac being the most dynamic mark (Saeed et al., 2014).
Among the genes displaying increased H3K27ac, we identified enrichment in genes encoding for proteins involved in cellular signaling as well as inflammation: cytokines and chemokines, the EGFR and VEGF pathways.
In addition, the PI3K/AKT pathway is also shown to be upregulated, and this is supported by the recent studies demonstrating its involvement in the induction of trained immunity by both BCG and β-glucan (Arts et al., 2016, Cheng et al., 2014).
The epigenetic profile induced by BCG shows close similarities to the trained immunity induced by β-glucan in vitro (Saeed et al., 2014), suggesting there is a core trained immunity response induced by different stimuli.
The epigenetic reprogramming of monocytes induced by BCG vaccination is accompanied by significantly altered responses of innate immune cells.
This process is clearly illustrated here by higher pro-inflammatory cytokine production (TNFα, IL-1β, IL-6) of PBMCs from BCG-vaccinated volunteers, compared to placebo-treated individuals.
We have in this case opted for PBMC stimulation (and not highly purified monocytes) to mimic as closely as possible the in vivo clinical situation.
In addition, while the epigenetic changes observed in highly purified monocytes are indicative of the induction of trained immunity programs by BCG vaccination, the functional consequences measured in PBMCs may be also influenced by the presence of lymphocytes.
Differences in cytokine production after BCG vaccination were seen in earlier studies both in children (Jensen et al., 2015) and in adults (Kleinnijenhuis et al., 2012).
The amplitude of the cytokine responses indicative of trained immunity induction in the current study is somewhat less strong compared to previous studies, probably due to BCG batch differences (Biering-Sørensen et al., 2015).
Unequivocally, demonstration of a correlation between the induction of trained immunity and protection against a controlled infection in humans has been outstanding.
Our proof-of-principle study showed that BCG vaccination protects against a subsequent, non-related viral infection in a human controlled model of YFV infection.
This model has the advantage of the use of a live virus, which has no deleterious effects on the host and which can be quantitatively monitored by PCR.
The demonstration of a significantly lower peak of viremia in BCG-vaccinated individuals compared to individuals injected with placebo clearly demonstrates the capacity of BCG to protect against non-related infections in humans.
The lower viremia was also accompanied by lower circulating inflammatory mediators in the BCG-vaccinated individuals.
This result is expected, as the high cytokine production capacity upon BCG vaccination will lead to a rapid local antimicrobial response and subsequent elimination of the pathogen, thereby preventing a systemic reaction and high levels of circulating cytokines (Netea et al., 2003, van der Poll et al., 2017).
Importantly, despite the lower virus load, BCG vaccination did not affect generation of protective anti-yellow fever antibodies and did not affect the specific effect of YFV.
This suggests that BCG might improve the antigen-presenting capacity and adaptive responses, in line with a study showing beneficial effects on the response to influenza vaccine (Leentjens et al., 2015).
The observed effects that are indicative of trained immunity induction are clinically very relevant, as epidemiological studies have shown that BCG vaccination results in lower all-cause mortality in the first month after birth (Aaby et al., 2011, Roth et al., 2004).
However, long-term studies with broader analysis of monocyte function are warranted in order to determine the duration of the BCG effect.
Importantly, significant differences in epigenetic markers were apparent on monocytes from BCG responders (with low yellow fever viremia) and BCG non-responders (with higher viremia).
H3K27ac at the level of the NOD2 receptor (which is essential for the induction of BCG-induced trained immunity [Kleinnijenhuis et al., 2012]) correlated best with response to BCG.
Additional immune pathways related to cytokine production and innate immune responses were shown to be of importance.
One of the most remarkable observations was made when the correlates of protection against viremia were investigated.
It was previously suggested that development of a scar after BCG vaccination could be used as a marker for non-specific effects and childhood survival (Garly et al., 2003, Roth et al., 2005).
However, all BCG-vaccinated volunteers developed a scar with a comparable size (0.5–0.7 cm), and we observed no correlation of this with protection.
In addition, the fold change of heterologous T cell responses (IFNγ, IL-17, and IL-22 production) upon BCG vaccination did not show any correlation with the YFV viremia.
In contrast, the post-BCG fold increase of IL-1β production strongly correlated with lower viremia after YFV administration, while production of other monocyte-derived cytokines showed less correlation.
This suggests that induction of IL-1β production stimulated by non-related pathogens, which is an indicator of trained immunity responses, rather than adaptive cellular responses as assessed by IFNγ production induced by specific stimuli, is responsible for the protection induced by BCG against the viral infection.
The fact that the increase in the capacity of innate immune cells to release IL-1β is the stronger correlate of protection suggests that this cytokine is a crucial component of trained immunity.
This finding is highly relevant due to the direct protective effect of higher IL-1β levels in viral infections (Azuma et al., 1992, Iida et al., 1989, Sergerie et al., 2007), although other mechanisms are possible as well.
The role of IL-1β is supported by complementary genetic and immunological studies.
By studying in vitro training of monocytes induced by BCG in cells isolated from healthy volunteers of the 200FG cohort (Li et al., 2016), we demonstrate that genetic polymorphisms in the gene encoding for IL-1β, but also other genes of the IL-1 pathway such as the inflammasome component ASC/PYCARD and genes encoding for the IL-1 and IL-18 receptors, are associated with the magnitude of the individual trained immunity response induced by BCG.
In follow-up experiments, we found that IL-1β itself can induce trained immunity, and that this effect is accompanied by epigenetic changes at the level of histone methylation (H3K4me3 and H3K9me3).
This may explain earlier studies that have shown that administration of one dose of IL-1β to mice before infection can protect against lethal bacterial and fungal sepsis (van der Meer et al., 1988).
In addition, IL-1β is well known to exert strong effects on myelopoiesis (Dinarello, 2002), and one may hypothesize that IL-1β represents the endogenous mediator between the peripheral stimulation of monocytes and macrophages by BCG and long-term functional reprogramming at the level of bone marrow progenitors.
Indeed, accompanying studies from the International Trained Immunity Consortium (INTRIM) in an upcoming issue of Cell demonstrate that effects at the level of myeloid cell progenitors are crucial for the induction of trained immunity by β-glucan (Mitroulis et al., 2018), BCG (Kaufmann et al., 2018), and Western-type diet (Christ et al., 2018) and identify IL-1β and the inflammasome pathway as key mediators of this process (Mitroulis et al., 2018, Christ et al., 2018).
Mitroulis et al. show that the adaptations in hematopoietic stem and progenitor cells (HSPCs) are associated with an increase in IL-1β (and not other cytokines) in the bone marrow fluid, while inhibition of IL-1β prevented the β-glucan-dependent expansion of HSPCs and the effects on myeloid cell progenitors. Similarly, Christ et al. demonstrate that a Western-type diet is able to induce trained immunity in mice, and IL-1 plays an important role in this effect.
Thus, Western-type diet did not induce trained immunity and myeloid progenitor reprogramming in NLRP3 knockout mice that have defects in IL-1β processing.
Similarly, genetic analysis revealed that polymorphisms in NLRP3 influenced oxLDL-induced trained immunity in human cells.
As dysregulated IL-1β production plays a pivotal role in the etiology of many autoinflammatory diseases, the capacity of IL-1β to induce trained immunity may represent a central event in the pathophysiology of these disorders.
This needs further investigation, as it may represent a therapeutic target for these diseases. Indeed, the accompanying study by Bekkering et al. of the INTRIM Consortium shows a dysregulated trained immunity phenotype in monocytes isolated from patients with hyper IgD syndrome (Bekkering et al., 2018), a well-known autoinflammatory syndrome.
In conclusion, in the present study we report the broad epigenetic program induced by BCG vaccination in humans, which results in increased activation of circulating monocytes. Furthermore, we show that BCG vaccination can protect against a non-related viral infection in an experimental model of human infection and report that IL-1β-mediated responses, which are indicative of the induction of trained immunity, are the most reliable correlate of this protection.
Additional genetic and immunological validation studies demonstrate that IL-1 pathway is crucial for an efficient induction of trained immunity in humans, and this may have important implications for both vaccination and the pathophysiology of autoinflammatory diseases.