Mycobacterium bovis bacillus Calmette–Guérin (BCG) is currently the only licensed vaccine for tuberculosis (TB). Although its protective efficacy against adult pulmonary TB is variable (1–3), BCG affords more reliable protection against childhood forms of the disease when administered to infants (4). The immune mechanisms responsible for this protection are not fully understood, although many studies have described BCG-induced, antigen-specific immune responses that may play a part (5–12).

Evidence also exists of a beneficial effect for the BCG vaccine against several different diseases and outcomes other than TB, including as a therapeutic agent against bladder cancer when instilled directly into the bladder (13) and protection against all-cause mortality in low birth weight infants (14). These effects are heterologous (non-specific) as they do not rely on mycobacteria-specific adaptive immune responses. Exactly how BCG mediates these effects is not clear, although different pathways are probably involved. Non-specific protective effects against mortality in low birth weight infants are a result of resistance to pathogenic microorganisms that are responsible for death due to neonatal sepsis and respiratory infections (15, 16). The immune responses mediating this effect could be either heterologous T-cells (i.e., T-cells induced by an epitope from one organism but with cross-reactivity to others) or the ability of BCG to potentiate the responsiveness of the innate immune system to later infections: a biological process termed trained innate immunity which represents a de facto innate immune memory (17–19). Such a phenomenon is thought to have existed for millions of years as acquired systemic resistance in plants and also in invertebrates, neither of which have adaptive immune systems (17). Several studies have revealed enhancements of the neonatal innate immune response to Gram-positive and Gram-negative infections following previous encounters, mediated by different toll-like receptor pathways (20) or initiated by in utero inflammatory conditions such as histologic chorioamnionitis (21). Studies in adults show that trained innate cells, including monocytes and natural killer (NK) cells, appear to be epigenetically and metabolically reprogrammed to produce increased amounts of pro-inflammatory cytokines and display higher levels of surface activation markers in response to restimulation with toll-like receptor ligands or different whole microorganisms (22–25).

If trained immunity is to provide a mechanism of action for the non-specific protective effects of BCG, evidence of trained immunity in infants or infant innate cells is needed. The immune system of the newborn infant differs from that of the adult in its constitution as well as in its propensity to respond to different stimuli; differences that reflect the unique physiological challenge of transitioning from the intrauterine environment to the outside world where some microorganisms are beneficial commensals and some are life-threatening pathogens (26, 27). Newborn innate immune responses are characterized by a reduced capacity for pro-inflammatory cytokines and dendritic cell differentiation and activation, but a greater propensity to produce regulatory cytokines (28, 29). More data are needed on how these differences impact upon the generation of trained innate immunity by BCG.

Our aim in this study was to probe the infant immune response to mycobacteria-unrelated stimuli following BCG vaccination for potential mediators of trained immunity. We used a multiplex bead array approach for the detection of secreted cytokines and chemokines from diluted whole blood following stimulation with a panel of innate stimuli. The rationale for this was to maximize the variety of immune cells available for stimulation as well as the potential to detect a broad array of soluble mediators. Our data provide a description of previously unreported profiles of trained immunity in infants and a role for activated NK cells.

Trained innate immunity has, to date, been demonstrated and characterized largely in adults. Isolated PBMC released more TNFα, IL-6, and IL-1β when exposed to non-specific innate stimuli 3 months after BCG vaccination in adults, and monocytes displayed higher levels of activation markers (24). Mechanisms of trained immunity such as underlying epigenetic regulation and shifts in the metabolic pathways used have similarly been revealed in vaccinated adults and in adult cellular models in vitro (22, 32). If trained innate immunity is to provide a causal link between BCG vaccination of infants and non-specific protection against mortality due to infectious diseases other than those with a mycobacterial etiology (15, 16), then a detailed description of the phenomenon in infants and infant cells is needed.

In this paper, we have profiled the infant whole blood cytokine and chemokine secretome of trained immunity following BCG vaccination using a panel of non-specific innate stimuli and reveal a differentially expressed signature. A total of 11 analytes were overexpressed and 6 under-expressed in vaccinated infants, with different innate stimuli inducing distinct combinations of response elements: for example, the Pam3Cys-induced response included the overexpression of 10 analytes, while S. aureus only induced three analytes. Four analytes were overexpressed and four under-expressed in response to C. albicans; LPS and E. coli each induced one overexpressed and two under-expressed analytes. These distinctions most likely reflect the different characteristics of the stimuli involved. Two of them, LPS and Pam3Cys, are defined biochemical compounds that interact with specific pattern recognition receptors: TLR4 and TLR2/1 heterodimer, respectively. C. albicans, S. aureus, and E. coli are all HK preparations of microorganisms containing complex mixtures of pathogen-associated molecular patterns that will interact with an array of different pathogen recognition receptors or other innate antigen receptors. It has been suggested that different microorganisms act on multiple, distinct TLRs simultaneously in a manner likened to the playing of a unique “chord” on a “molecular piano” (17) and the different response profiles generated by different stimuli described here may be an illustration of that effect.

The pathways by which BCG mediates its training effect in vivo are yet to be fully elucidated, although a role for the mycobacterial cell wall component muramyl dipeptide, acting via the NOD2 receptor, has been described (24). β-glucan, another activator of trained immunity mediates its effects via a dectin-1/Raf1-dependant pathway (33), yet the role of dectin-1 in recognizing mycobacteria is probably minor at most. The emerging picture is one in which the recall of a trained immune response may be restimulated via different receptors to those that prime that same response and that the cytokine/chemokine signature of that recall is determined by the receptors involved. Although little is known at this time about the mechanisms responsible, this should be a focus of future studies.

Despite the differences described above, we described here a biosignature of BCG-induced trained innate immunity in infants comprising increases in EGF, IL-6, and PDGF-AB/BB, all of which were apparent in responses to three different innate stimuli. This is different to the characteristic signature of trained innate immunity in adults, which includes TNFα and IL-1β as well as IL-6. It must be however underlined that in adults not all the cytokines reported here were measured, and some of them may be upregulated as well in older individuals. The absence of canonical innate cytokines such as TNFα in this infant biosignature may reflect the maturation of the infant innate immune response, whereby the ability to make certain cytokines develop over the first 9 months of life (34). Despite the most extensive response profile, Mtb lystae stimulation did not upregulate EGF or PDGF-AB/BB. This may reflect the involvement of a broader array of pattern recognition receptors in response to Mtb lystae with downstream effects that manifest themselves as a distinct pattern of upregulated and downregulated markers. These observations illustrate the importance of an unbiased, multiplex approach to analyzing these responses. A more directed study focused on TNFα, IL-6, and IL-1β would have underestimated the effect of trained immunity in infants and would have missed the previously unreported markers EGF and PDGF-AB/BB, as well as the finding that C. albicans-induced trained immunity recall in BCG-vaccinated infants involves reduced IL-2, IL-13, IL-17, and IP-10 responses. It is interesting that IL-17 responses to C. albicans are reduced in vaccinated infants. Th-17 cells are known to be important in immune responses to C. albicans (35). BCG vaccination has been shown to enhance non-specific (innate-mediated) protection against candidiasis in SCID mice (24) and adult BCG vaccination activates heterologous Th-17 responses specific for non-mycobacterial ligands including C. albicans (31). Our data suggest that BCG is exerting a different influence on the infant immune response, steering it away from a bias toward Th-17 development that is known to exist in early life (36). Detailed analyses of epigenetic and transcriptional programs of training or tolerance induction have revealed upregulation and downregulation of genes that characterize these responses hence it is unsurprising that the profiles we report here involve both upregulation and downregulation of different cytokines and chemokines associated with both T-cell and inflammatory monocyte responses to stimuli such as C. albicans as well as to E. coli and LPS which have both been associated with the induction of tolerance (37, 38).

In contrast to our findings here, another report of infant BCG-associated, non-specific cytokine responses described increases in TNFα and IL-1β as well as in IL-6 in response to Pam3Cys (39). Differences between study designs that might explain this difference are that Jensen et al. carried out their study in a low-income country and in low birth weight infants. In addition, the genetic backgrounds of the populations studied were different. Their sampling time point of 4 weeks was earlier than that used here and there were some methodological differences in their whole blood assay. An interesting possible explanation is that, unlike Jensen et al., infants in our study had received DTP vaccination through the national vaccination program by the time of sampling which is thought to negatively affect the positive non-specific effects of BCG (40). Finally, compared to Jensen et al., our study was relatively small and exploratory and was possibly underpowered to detect small differences in cytokine responses. It should be noted that in agreement with our findings and those of others (7), Jensen et al. found no BCG-associated increase in TNFα in response to non-specific LPS stimulation.

The involvement of monocytes in adult trained immunity has been demonstrated using stimulating ligands that interact with receptors known to be found on monocytes and more directly by the demonstration that markers of monocyte activation and differentiation are upregulated following training (24). It is also possible to isolate adult blood monocytes and train them directly with BCG and β-glucan in vitro (32). Although the assay described in this paper was designed primarily to look at secreted cytokines and chemokines, it was possible to harvest PBLs and to examine monocytes and NK cells for evidence of BCG vaccination-associated trained innate activation. We examined four markers of monocyte activation, but found no significant differences in their expression following non-specific stimulation of samples from vaccinated and unvaccinated infants. Stimulation with Mtb lystae, which also induced increases in 27 cytokines and chemokines in vaccinated infants, did increase levels of CD11b and CD207 on the surface of monocytes. Unlike responses induced by non-specific ligands in these experiments, it is likely that antigen-specific T-cell help is involved in these changes in monocyte activation. It is interesting that monocytes from BCG-vaccinated infants show increased expression of CD207 (mannose receptor), as this is usually found only on certain populations of differentiated macrophages and dendritic cells (41).

Natural killer cells from BCG-vaccinated infants displayed increased levels of the activation marker CD69 in response to Pam3Cys, as well as to Mtb lystae. Previous work has described long-lived mycobacteria-specific NK cells following infant BCG as well as a role for NK cells in BCG-induced trained immunity in adults and in BCG-vaccinated SCID mice which show increased resistance to C. albicans infection (25, 42). However, this is the first time their involvement in trained immunity to heterologous stimuli in infants has been described. NK cells express the TLR2 receptor via which they are reported to interact with mycobacteria (43, 44). NK cells are also activated in response to type 1 cytokines (45). Although we cannot determine the pathway of NK cell activation from the current data, we did observe a correlation between CD69 expression, IL-12p40 secretion, and IL-10 secretion in response to Pam3Cys. IL-12 is known to be a soluble activator of NK cells, while a role for IL-10 is less obvious. A limitation of this study was the small volume of blood available from infants which meant it was not possible to identify the cell type producing the cytokines and chemokines described. Clearly it would be interesting to determine whether the NK responses described above or other populations were a source, either by depletion studies or intracellular cytokine staining and future studies should address this. Additionally, as well as stimulation of the TLR2/1 heterodimer using Pam3Cys, it would be useful in future to determine more precisely the role of TLR2 either in homodimeric form or as a heterodimer with TLR6 also using specific ligands for these receptors.

This exploratory study was not designed to investigate the heterologous effects of infant BCG vaccination on clinical outcomes; however, if the non-specific immune responses we describe are to provide a mechanism for these effects, a causal relationship will need to be demonstrated in a larger study. The setting of such a study will need to be carefully considered as to date, the results of clinical trials have been mixed. Heterologous, BCG-induced protection against clinical outcomes related to infectious disease has been observed in a low-income setting in West Africa (15, 46) but was not recapitulated in another trial in a high income, European setting (47, 48).

In conclusion, we have described a novel whole blood signature of BCG-induced, trained innate immunity in infants that includes secretion of EGF, IL-6, and PDGF-AB/BB and involves activated NK cells. We also show distinct patterns of cytokine and chemokine release in response to different innate ligands. The data show different patterns to previously published descriptions of BCG-induced trained innate immunity in adults and low birth weight infants and suggest that more, larger scale studies of this effect in different populations are required for a more complete understanding.

This study was carried out in accordance with the recommendations of the National Research Ethics Service Committee London-East and the ethics committee of the London School of Hygiene and Tropical Medicine with written informed consent from all subjects. All subjects gave written informed consent in accordance with the Declaration of Helsinki. The protocol was approved by the National Research Ethics Service Committee London-East and the ethics committee of the London School of Hygiene and Tropical Medicine.

SS performed the experiments, was responsible for the analysis and interpretation of data, and wrote the manuscript. JK and MN produced the panel of stimuli used in experiments. SS, JK, MN, and HD all contributed to the conception and design of the work, critically read and contributed to revisions of the manuscript, approved the final version of the manuscript, and agreed to be accountable for all aspects of the work including its accuracy and integrity.

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.