Key words

1 Introduction

Systems vaccinology (SV) is a nascent science which has emerged as a variant of systems biology and aims to understand the effect of vaccines on the entire host system. In contrast to systems biology , SV is therefore heavily application focused, and possibly only in that respect different from the more generally used term systems immunology . Like all areas of systems biology, the aim of SV is to overcome reductionist thinking and generate a more comprehensive, dynamic, and in a way more realistic understanding of the interaction of components in an organism, in this specific context of reactions induced by vaccines or factors influencing efficacy of vaccines. While this is not a new concept, viewing the host as a system interacting with the environment (also in many cases a set of interacting systems) allows identification of signatures and interactions critical for defining or achieving states critical for success of a therapy, drug, or vaccine . These signatures may not be evident from a reductionist point of view, at least not without complete deconvolution of all possible components. Being embedded in a complex system, understanding host components and their interaction with environment in a single picture is critical for rationally designing drugs including structure-based vaccines . This is particularly important in the case of chronic infections where the system constituting an intruding pathogen establishes a continuous but unfortunately not mutually beneficial relationship with the host system. Understanding host/pathogen interfaces will therefore be another important area within systems vaccinology and may very likely break the boundary to other areas of pharmacology to further expose and target this host/pathogen interface. Chronic infections are typically established by a pathogen by mechanisms which make host defenses at least partially ineffective by remodeling host immunity as well as by immunological stealthiness. Likewise, cancer vaccines (with few exceptions) to date have not been very successful. Understanding the complex interaction of cancers with their environment (including tumor stroma) employing the concept of SV will facilitate effective cancer vaccines. While vaccines have been tremendously successful, the modes of action of a number of commonly used vaccines are not well understood. Infectious diseases, including HIV, malaria, tuberculosis, and dengue fever, have proven to be at least partially resistant to traditional vaccination approaches. Systems vaccinology offers powerful tools to monitor a multitude of data types sampled in parallel through omics technologies to reconstruct a comprehensive picture of processes leading to success or failure of a vaccine or a vaccine prototype. As recently reviewed by Pulendran, Rappuoli, and others, SV therefore primarily offers a deeper understanding of processes leading to or prohibiting protective immunity and establishes a basis for vaccine developers to overcome old limitations [1, 2]. This approach of understanding vaccines will become even more complex as sources of variation between immune response s become more evident and tangible, specifically genetic host diversity, environmental and psychological factors, impacts of the microbiome , and socioeconomic factors including associated disorders such as obesity, diabetes, and those resulting from malnutrition. Systems vaccinology offers a way to gain deeper understanding of the ongoing pathomechanisms by comparing to successful immune responses, thereby leading to the development of new immunization strategies for diseases that do not yet have a commercially successful vaccine. Successful vaccines that induce the immune system can be modeled based on expression signatures or molecular processes that clear pathogens during an infection. Data from entire populations as well as nonresponders could help in designing vaccines. The type and intensity of immune reactions can among other things be effectively modulated by pattern recognition receptors (PRRs) including Toll-like receptors (TLRs) and nucleotide-binding oligomerization domain (NOD) proteins, which constitute some of the most effective adjuvant targets known. Understanding the immune-evasive mechanisms as well as the latest immunological insights and mechanisms of novel adjuvants will lead to reprogram ineffective to effective immunity.

Some currently pursued and further perceivable applications of systems vaccinology are listed below:

  1. 1.

    Predict immunogenicity and optimal protectivity of vaccines in an individual based on host samples taken before or after vaccination. Applications include stratification of patients for clinical trials, early identification of nonresponders which may allow application of alternative medications, and risk assessment regarding adverse effects.

  2. 2.

    During vaccine research, taking comprehensive immunological snapshots should speed up development, reduce the risk of failures in later clinical trials, and improve understanding of immunity in risk groups including the very young and the elderly. Specific identification of novel formulations and adjuvants, ex vivo (peripheral blood mononuclear cell (PBMC)) optimization of dosage, relative and absolute adjuvant content, and better understanding of the relationship of immunity with metabolism and neuronal system are required.

  3. 3.

    Another goal is to understand the interface of vaccines with complex phenomena like cancer and autoimmunity, including autoantibody levels and antibody subclasses. Understanding the potential of vaccines to modify diseases with inflammatory components such as neoplastic disorders, stroke, and coronary heart diseases may have drastic beneficial effects on public health . This extends to the idea to use self-associated molecular patterns (SAMPs) to reduce inflammation, an option which seems tangible due to insights into the glycobiology of antibodies including anti-inflammatory effects of targeted Fc receptors.

2 Understanding and Prediction of Vaccine Responses

The most frequent application of systems vaccinology to date has been the definition of critical parameters associated with immunogenicity and optimal protectivity of a vaccine . The aim of a number of studies has been to define systems signatures, specifically based on gene expression analysis, which can be used as surrogates for vaccine efficacy. Application of these signatures would allow:

  1. 1.

    Evaluation and optimization of vaccine candidates including choice of adjuvants or dosage based on observation of signatures in PBMCs in vitro.

  2. 2.

    Early identification of nonresponders after vaccination.

Meanwhile the concept has been extended to include features of the host before vaccination. A number of molecular and other properties have been identified to allow estimation of the vaccine response before treating a subject, which has so far been aimed for two related practical applications:

  1. 1.

    Stratification of patients for clinical trials based on likelihood to react to a vaccine .

  2. 2.

    Early identification of potential nonresponders before vaccination, which would allow selection of different vaccine products, dosage, or administration regimens.

Both the latter options are forms of personalized medicine. A third related option may be the early identification of candidates for adverse side effects including rare conditions such as narcolepsy.

The highly effective yellow fever vaccine , YF-17D, was the first in a study of systems vaccinology where vaccinology and immunology were combined [3–6]. The initial study by Querec et al. succeeded to define and verify early post-vaccination gene expression signatures which predicted YF-17D adaptive immune response s, specifically CD8+ (cytotoxic) T-cell responses with up to 90 % accuracy and the neutralizing antibody response with up to 100 % accuracy depending on evaluation procedure. This notably only requires models involving two to three parameters each. Given the wide use of this vaccine it may also be a good candidate for better understanding and prediction of adverse side effects including rare serious adverse effects [7–9]. Technologically this initial study has been shaping the conceptual approach to systems vaccinology until now. The authors combined FACS characterization of immune cells from patients sampled after vaccination in a time series, along with a multiplex panel of cytokines and gene expression analysis of PBMCs using microarrays. Gene expression data were analyzed for statistically significant changes using analysis of variance (ANOVA) of the entire time series of fold changes. Results were corrected for multiple testing resulting in a set of 65 genes putatively differentially regulated in PBMCs upon vaccination. RT-PCR was used to validate differential expression of a subset of the indicated genes. For biological interpretation of microarray analysis results, differentially regulated genes were then subjected to gene enrichment analysis using DAVID to identify enriched pathways and modules [10]. In addition transcription factor-binding sites statistically overrepresented in the promoters of these genes were identified using TOUCAN [11]. Obtained immune reactions were characterized by epitope -specific T-cell assays and determining neutralizing antibody titers. Interestingly the 65 differentially regulated genes were not useful (or insufficient) to predict magnitude of CD8+ T-cell responses. For this purpose expression fold changes of individual genes at day 3 and separately at day 7 were individually tested for significant correlation with later CD8+ T-cell response and neutralizing antibody response. In addition to PCA analysis which indicated a segregation of high and low CD8+ responders, average linkage hierarchical clustering analysis was used and confirmed segregation of these two groups, followed by feature selection, model building, cross-validation, and further independent validation.

Based on samples before and after vaccination Nakaya et al. have demonstrated that signatures of protectivity between live (LAIV) and inactivated (TIV) influenza vaccine s vary where the authors specifically stress IFN-related genes as differentiating signature upon LAIV administration [12]. Similar to Querec et al. this study aimed to predict vaccine efficacy based on samples taken a few days after vaccination to allow prediction of hemagglutinin -inhibiting (HAI) titers a month later and specifically to allow identification of vaccine nonresponders as early a possible. While this study technologically also focused on gene expression analysis (microarrays and RT-PCR) and FACS they individually determined gene expression for PBMC subsets separated by FACS. This approach is highly reflective of the functionally differentiated nature of PBMCs and can lead to the identification of marker signatures otherwise invisible in bulk analysis due to expression primarily in critical but low-abundance subpopulations, demonstrated here by plasmacytoid dendritic cells (pDC). More cost- and logistically affordable analysis methods also capable of differentiating cell types have been proposed, specifically a meta-analysis procedure by Nakaya et al. and also statistical deconvolution using cell type-specific significance analysis of microarrays (CSSAM) [13]. In the case of Nakaya et al. meta-analysis suggested a TIV-specific pattern of highly expressed genes in antibody-secreting cells (ASCs), whereas cell types implicated in LAIV analysis were specifically T-cells and monocytes. These results show the importance of different cell types during analysis and how they modulate protective mechanisms during vaccination. A similar study of TIV without differentiating cell types was conducted by Bucasas et al. and also indicated a distinct gene expression signature comprising 494 genes segregating high and low vaccine responders [14]. Likewise, in a study conducted by Vahey et al. for prediction of protectivity of a malaria vaccine , the immunoproteasome and apoptosis were indicated as marker processes, depending on time after vaccination [15].

Furman et al. have strived to define markers of immunological health, where their idea was to use the intensity of immune reactions against an influenza vaccine as a measure of health [16, 17]. Notably, they identified nine features available before vaccination which predicted vaccine response with 84 % accuracy, plus a number of other parameters which also were individually predictive to some degree but did not further improve the model. Overall analyzed data included patient age, whole-blood gene expression profiles, peptide-specific anti-influenza antibody titer measured by peptide microarrays, 50 cytokines and chemokines, and typing to a resolution of 15 immune cell subtypes using FACS. The authors found that taken alone, age is the most informative predictor of HAI titers, where increased age correlated to reduced titers. As previously published by others, pre-existing HAI titers (for example stemming from previous vaccination or natural infection) are negatively associated with influenza vaccine-generated HAI titers, reportedly resulting from limited dendritic cell (DC) antigen presentation caused by pre-existing memory T-cells via natural killer (NK) cells. Furman et al. were able to specify a number of peptides which are predictive of this effect before vaccination and those that provide independent information to predict vaccine efficacy. In addition, whole-blood gene expression analysis followed by analysis of gene modules proposed a number of gene sets where expression was negatively correlated with post-vaccination HAI titers, and one associated with apoptosis which was positively correlated. The identification of apoptosis as critical parameter is reminiscent of the results described by Vahey et al. in malaria. The relevant information for predicting vaccine response was also present in levels of soluble Fas ligand (sFasL) and IL-12p40, and frequency of central memory CD4+ and effector memory CD8+ T-cells. From a number of perspectives also the setup of this study can be seen as exemplary for the identification of predictive models of vaccine efficacy, predominantly because of the number of omics techniques applied. Furman et al. observed (epitope specific) antibody titers based on peptide arrays, gene expression analysis of blood cells, and FACS-based immune cell subtyping. Independently, Tsang et al. have compared pre- and post-vaccination PBMCs using FACS and could show that pre-vaccination cell population frequencies alone are predictive of antibody response for TIV [18]. FACS-based discrimination of immune cell types followed by microarray analysis and cytokine panels has led to identification of PD-1+CXCR5+ CD4+ T-cell numbers as indicators of emergence of broadly HIV-neutralizing antibodies. This population is similar to CD4+ T follicular helper (Tfh) cells critical for B-cell maturation in germinal centers (GC), indicating the potential to characterize immune systems by available cell repertoire and opening strategies to stimulate specific populations for vaccine effect [19].

At least one of the possible ways to define effective vaccines is to learn from immune response s to natural infections where the pathogens are effectively cleared. Based on the data the correlate of protection may be identified, optimally reducing the need for challenge studies. At least in a mouse model of influenza, the peptide arraybased immuno-signatures have been identified in natural infection which allows differentiation of protective and non-protective vaccine immune responses by pattern [20]. For this chip “long, pseudorandom, nonnatural” peptides are used which may, at least theoretically, also allow identification of antibodies binding posttranslational modifications (including glycosylations) through peptide molecular mimicry. This may be an interesting template in diseases where effective vaccines exist or natural immunity is protective to serve as a template. Another important factor in predicting vaccine responses accessible to systems technologies includes SNP chips and next-generation sequencing of the host genetic background. Although this factor is very difficult to differentiate from environmental factors such as the microbiome , previous studies in twins and families have suggested heritability in vaccine responses to range between 39 and 90 % depending on vaccine and degree of hereditary relationship [21–28]. Interestingly, antibody levels against a multitude of common pathogens seem to be heritable. Similarly, in autoimmune disorders genetic background along with environment plays a crucial role in establishing specific immune phenotypes [29]. It is therefore reasonable to include high-throughput screening for host genetic markers into systems vaccinology procedures for prediction of vaccine efficacy or possibly adaptation of dosage. An approach based on SNP and microarray gene expression analysis for identification of genetic markers impacting vaccine response has been implemented by Franco et al. suggesting 20 genes that show evidence of significant genotype expression association related to TIV [30].

However while proteomics, metabolomics, and glycomics might contribute beneficially to the development of more robust or less resource-intensive models, the published studies show a critical requirement of systems vaccinology (and systems biology): a large number of high-quality samples, optimally leading to significantly more samples than measured parameters (which is admittedly unlikely), or possibly integrated parameters as seen in gene sets versus individual gene expression values. In vaccines where T-cell responses play a critical role, dimensionality can be reasonably further increased by ELISPOT or comparable T-cell epitope -specific data equivalent to humoral peptide array assays. While powerful feature selection procedures exist, the factor of dimensionality needs to be considered and robustness of models can be enhanced with larger numbers of samples, specifically if biologically alternative possibilities for achieving complex immune phenotypes such as protectivity exist. Typically numerous feature selection methods will be compared, where some like DAMIP, ClaNC, or elastic nets have proven specifically useful for high-dimensional biological data [31–34].

While the initial hope of SV has been to identify general profiles (signatures) indicative of protectivity, signatures identified so far are predictive of immune response s for a specific vaccine or vaccine type sharing a mode of action (or adjuvanting), allowing to predict success or insufficient protection in individuals [35, 36]. It can also be expected that a tighter integration of vaccinology and immunology will lead to a feedback with related disciplines, specifically research into autoimmunity where delineation of cofactors for development of immune phenotypes including gender, hormone status, or infections has a long history [37].

3 Vaccines and the Role of Pre-existing Immunity

Immune systems are typically not neutral as they have previously encountered foreign agents and have built tolerance to self-components. Furman et al. and others have shown that pre-existing titers against influenza hemagglutinin reduce effectiveness of the trivalent inactivated seasonal influenza vaccine (TIV) in terms of total HA-neutralizing (HAI) titer achieved [38, 39]. This effect does not seem to exist for an influenza live vaccine (LAIV) [40].

Current evidence suggests possible detrimental role of low-affinity antibodies stemming from previous infections with related pathogens as demonstrated in Dengue virus infection. Infection by a single serotype is in most cases harmless, or comparable to infection as in common cold. However, infection by three or four serotypes frequently leads to a hemorrhagic syndrome, inducing cross-reactive T-cells and (low-affinity) antibody-mediated enhancement (ADE) [41–43].

An example where pre-existing immunity against the vaccine vector increases pathogen acquisition rates is Merck’s MRKAd5/HIV which is highly immunogenic but non-efficacious. This HIV-1 vaccine uses an inactivated adenovirus serotype 5 (Ad5) vaccine vector and seems to induce high HIV-1 infection rates in Ad5 seropositive individuals [44]. While reason behind this effect has been suggested to be at least in part antibody-mediated uptake leading to increased dendritic cell activation, recent systems analysis by Zak et al. suggests that pre-existing Ad5-neutralizing antibodies effectively reduce dose of vaccine and hence immunogenicity [45, 46].

It is also well established that stimulating the immune system in a way as to promote an immune arm unsuitable for pathogen clearance can be an immuno-evasive strategy for pathogens, exemplarily demonstrated by the Th1 versus Th2 immune signatures seen in leprosy and the role of IL-10 in Epstein-Barr virus (EBV) infection [47, 48]. The induced state of the immune system and degree of cross-reactivity of the adaptive immune system can be expected to severely shape efficacy and precise nature the host responds to secondary natural infections and also to vaccines [49–52]. Better understanding of the precise interaction of superinfections and interaction of complex immune phenotypes with vaccines may contribute substantially to prediction of inter-patient variability in vaccine responses. This calls for advances in immunology , molecular biology, and systems vaccinology.

4 Microbiota, Chronic Infections, and Vaccines

While the extent of interaction can be expected to go substantially beyond current knowledge, analysis has shown that gut microbiota are critical for achieving potent immunity using inactivated influenza vaccine s. Toll-like receptor 5 (TLR5)-mediated pathway is critical for vaccine efficacy where gut microbiota provide stimuli for development of plasma cells ultimately impacting antibody production, while live vaccine s and adjuvanted vaccines may not share this dependency. Similarly, the importance of intestinal flora composition has been demonstrated for TLR7-stimulated development of inflammasomes in respiratory mucosa, where lack of TLR7 ligands leads to impaired immune response s to influenza [53, 54]. Likewise, it has recently been shown that pathogen -free mice are more sensitive to influenza challenge than other mice and that inflammation can be dampened via colonization with Streptococcus aureus in a TLR2-dependent way [55]. However, respiratory influenza infection can lead to gastroenteritis-like symptoms not via direct infection of gut epithelia, but rather due to a shift in gut microbiota leading to increase in Th17 cells in the small intestine and also enhanced IL15/IL17A production, an effect abolished by antibiotics [56]. Possibly adding some detail to this observation, Weber et al. conclude that IL17-producing thymocytes form a “first line of recognition” stimulated by cell wall components of diverse pathogenic and apathogenic bacteria, but that effector molecules such as IL-6 and IFN-γ determine transition to a pathological inflammation [57]. It has also been demonstrated that microbiota depletion impairs early innate immunity against the pathogen Klebsiella pneumoniae and that this state can be remedied by providing NOD-like (NLR) receptor ligands but not Toll-like receptor (TLR) ligands from the gastrointestinal tract, whereas NLR ligands from the upper respiratory tract were ineffective [58]. This highlights the systemic impact of local microbiota on immune responses and suggests a critical importance of microbiota derived pattern recognition receptor (PRR) ligands for establishing effective immunity. In summary it can be stated that current evidence shows heavy dependency of microbiota and microbiota composition on activity of the immune system and efficacy of at least some of the vaccines. Bacteria are not the only microorganisms modulating the immune system; the virome can support intestinal homeostasis comparable to bacterial commensals, presumably by providing equivalent stimuli [59, 60]. Similarly, fungal diversity and species composition may prove to be a critical extension also in other areas than chronic inflammatory disorders of the gut [61]. The terrible and disfiguring childhood disease Noma (cancrum oris) is currently thought to be caused by malnutrition and microflora dysbiosis [62]. There is also clear evidence that the choice of food impacts microbiome development and ultimately immune competence [63]. The interplay of human nutrition, gut microbiome, immune system development and competence, dysfunction, and vaccine efficacy is the focus of ongoing research and is now viewed as a very likely critical dimension of immunology and hence possibly also vaccinology [63–69]. The impact of microbiota and microbial diversity on vaccine efficacy in infants has recently been investigated in a small cohort by Huda et al. where they suggest probiotics for minimizing dysbiosis [70]. Of note in this context, while microbiota have emerged as an important immunological dimension, metagenomics has emerged as a powerful tool for analysis of the microbial community in an organism. Metagenomics could be used to identify and quantitate the gut microbiota of the fecal samples. Several other body (especially mucosal) surfaces are commonly covered by microbial communities; within the gastrointestinal system several distinct regions exist which contain microbiota of typically different composition. Moreover, the gel layer and luminal communities of gut microbiota have been shown to feature different population composition [71]. Finding ways to routinely access these spatial dimensions in health and disease may open yet another possible critical aspect for integration into the growing number of systems components regarding immunity and also vaccine effects. The gut glycome is another uninvestigated area presumably providing substantial immunologically relevant mass to the human body.

The distinction between a commensal and a pathogen can be difficult to draw, depending on the potential to be involved in disease. Examples of organism with potential impact on vaccinology are immune-distorting bacteria like Mycoplasma species which can cause diverse diseases in animals and humans [72] but are also nonobvious (and often unidentified) microbes of the natural microbiome [72, 73]. As multiple roles have been suggested for the human pathogen M. pneumoniae this may either mean a substantial under-appreciation of other causes of atypical pneumonia or otherwise of other factors contributing to the conversion from an asymptomatic infection to a severe disease. Mycoplasma species are also frequently associated with autoimmune diseases [74]. The mechanism by which Mycoplasmas modulate the immune system are not clear, and part of the reason is that these pathogens may contribute to a pro-inflammatory or otherwise immunologically biased environment rather than being clear-cut pathogens in the sense of Koch’s postulates. At least in chicken severe exacerbation of otherwise asymptomatic (avian) influenza infection with M. gallisepticum, a Mycoplasma phylogenetically close to M. pneumoniae, has been documented [75]. Several known interactions exist where these can lead to nonadditive exacerbation of other infections through intensified inflammatory responses. At least in Ureaplasma species the term pseudospecies has been used, as different isolates may vary greatly in their content of pathogenicity factors. In fact, from both a general health and a vaccine perspective it may be equally critical to consider the immuno-modulatory pathogenicity mechanisms available within a person’s microbiome along with specific bacterial species, as their combined effect may be very distinct or at least nonadditively amplified from individual factor contributions [76].

Another example of frequently observed chronic pathogen includes the highly prevalent immune distorting viruses of genus lymphocryptovirus comprising Epstein-Barr virus (EBV) and Cytomegalovirus (CMV); these cause lifelong infections, and EBV is known for its B-cell tropism. Both viruses can establish regulatory complex periods of latency. The effect of EBV and CMV infection versus age on immunity has recently been studied by Wang et al. where they differentiated age-dependent and -independent effects [77]. Specifically, decreased diversity of antibody repertoires with accumulation of memory B-cells and lower naive B-cell populations was associated with reduced vaccine efficacy in the elderly. They report that immune-globulin heavy chain (IGHV) mutation frequency increases upon infection with CMV, but not EBV. CMV infection tends to increase the proportion of highly mutated IgG and IgM regions, but not IgA or IgD. The effect of CMV on mutation rate is stronger with age, where the effect may stem from the proportion of CMV-specific clones. Age and EBV infection correlated with persistent clonal expansion, where very few clonal lineages (possibly derived from a single ancestor) tend to be overrepresented. In the study these expanded clones may be cases of monoclonal B-cell lymphocytosis (MBL), a lymphoproliferative disorder with some characteristics of CLL typically seen in the elderly.

The inflammatory status including degree of immune system activation can have significant impact on vaccine efficacy. Recently it was shown in a YF-17D (yellow fever vaccine) trial comparing vaccination efficacy of 50 volunteers in Lausanne (Switzerland) versus the same number in Entebbe (Uganda) that the latter produced less effective humoral and CD8+ responses. The authors negatively correlated the pre-existing activation level of CD8+ T-cells and B-cells as well as pro-inflammatory monocytes at the time of vaccination with this reduced response [78]. Admittedly it would also be interesting to know the cause of this inflammation, as the specific reason may affect the impact on vaccines. On the other hand the impact of pre-existing low-grade inflammatory conditions on vaccines is a recurring theme in the current review. In this context it is evident that determining protectivity profiles for vaccines is only one side of the coin. The other one is that the status of the vaccine recipient regarding inflammatory diseases, nutrition, and pre-existing immunity needs to be considered to understand inter-patient variability. Unfortunately the complex interaction of multiple clinical and subclinical infections is poorly understood. In the context of vaccines, inflammation and potential impact of chronically infecting pathogens and pathogen interactions need to be addressed.

5 Vaccines, Metabolism, Hormones, and the Nervous System

The interplay between metabolism and immune and nervous system is extensive and well beyond the scope of this review. However it is certainly beneficial to highlight some key concepts and current understanding regarding cross talk to show potential ramifications this may have on vaccine design . My aim here is to show factors similar to the preconditioning of chronically infecting pathogens impacting on the immunological environment within which a vaccine has to operate and which could be measured by a systems vaccinology approach. Indeed the relationship of inflammation and metabolic disorders has been extensively reviewed and is known to be very prominent [79]. Specifically the link and overlap between nutrient and pathogen -sensing mechanisms have been implicated in the development of inflammatory disorders. The biological rationale has been speculated to rest on the beneficial effect to coordinate short-time energy requirements during immune response with energy storage and metabolism, but the system has not evolved to deal with continuous nutrient surplus. Among others obesity, type 2 diabetes, cardiovascular disease, and certain neurologic disorders such as dementia and major depression have key low-grade (chronic) inflammatory components where this form of inflammation is distinct from acute inflammation involving swelling and pain. Low-grade inflammation is similar to classical inflammation on a molecular level, triggered by nutrients and metabolic surplus. It also turns out that immune components and metabolic organs may have evolved from the same source, as suggested by the fruit fly fat body which coordinates metabolic and pathogen-associated survival responses. Part of this link still seems to exist on pathway and physiological level. Examples are the lipopolysaccharide (LPS) receptor Toll-like receptor 4 (TLR4) which has been demonstrated to be directly activated by fatty acids and GCN2 which links dendritic cell autophagy and CD8+ cell antigen presentation (innate and adaptive immunity) to amino acid starvation [80]. TLR4 polymorphisms also have been linked to likelihood of developing type II diabetes in the Chinese population [81]. GCN2 was identified by Querec et al. within a systems vaccinology analysis as a factor frequently contained in YF-17D vaccine response efficacy determinants. Several immune receptors, including TLR4, TLR2, and NOD1, have been shown to play a role in adipocyte inflammation [82]. TNF-alpha and other pro-inflammatory cytokines are over-expressed in adipose tissue and can lead to insulin resistance [83]. In fact adipocytes share a number of similarities with lymphocytes including pathogen-sensing capabilities. Also, lipids are well known for their capacity to regulate metabolism and adaptive and innate immunity , at least partially through peroxisome-proliferator-activated receptor (PPAR) and liver X receptor (LXR) family transcription factors, repressing expression of inflammatory mediators [79]. Interestingly drugs acting through PPARγ like thiazolidinediones are potent insulin sensitizers, but inhibit TLR-mediated activation of dendritic cells [84, 85]. It has been shown that catecholamines and adipokines influence immunity, metabolism, and the central nervous system [86]. Catecholamines including dopamine, noradrenaline, and adrenaline are generated by a number of cell types and can mediate a multitude of neural, metabolic, and pro- and anti-inflammatory effects. Adipose tissues as key endocrine organs secrete adipokines and hormones which serve a number of functions, including being both pro- and anti-inflammatory immune mediators with possibly a role in neuroinflammation [87–89]. Adipocytes can be found in a number of tissues and depending on location they can have different secretory profiles of important factors including adipsin (factor D), TNF-alpha, IL-6, apelin, chemerin, resistin, MCP-1, PAI-1, RBP4, ghrelin, and visfatin. Obesity is also a well-studied factor for prediction of vaccine response. While presence of local or systemic low-grade inflammatory markers may be more informative than obesity itself it has been shown to enhance susceptibility to infections and reduce immune competence and vaccine efficacy [90–96]. Recent murine studies also suggest a specific role of B-cells and autoantibodies in obesity-related pathology [97]. Type II diabetes is now understood to contain a significant inflammatory component and has been associated with reduced efficacy of hepatitis B vaccination in China [98–101]. Another study in this context suggests that although vaccine titers are reduced this is not necessarily associated with reduced protectivity, indicating that type II diabetes alone may not significantly reduce vaccine-provided protection, at least not in children [102].

Some of the factors that may influence the immune system during vaccination include age and degree of obesity. Among the other factors known to influence the immune system and vaccine response, hormones and specifically the group of progestogens, hormone balance, and vitamins A and D are of particular relevance [103–105]. Hormones are considered to be the driver of immune differences between males and females, typically leading to weaker infection-related immune reactions in males and higher incidence of autoimmunity in females; intrinsically these observations are age dependent [106, 107]. Given current evidence, monitoring of hormone status should be a reasonable area for observation of future systems vaccinology studies and complements current procedures of monitoring serum proteins and other metabolites. Jensen et al. reviewed correlation of administered vaccines and vitamin A supplementation (VAS) including number of administrations. The data showed a significant sex-dependent difference (positive or negative) on mortality in monitored infants and VAS effect also depended on the location where it was administered (possibly because of ethnicities and associated genetic factors). Primarily, vitamin A supplementation is actively discussed as it is a WHO recommendation, and there is a differential effect on VAS-related vaccine responses between boys and girls [108, 109]. Independently it has also been reported that malnutrition overall has little effect on vaccine responses, suggesting that VAS administration at young age should be handled with care and sex dependency has to be further considered [110]. While oral but not parenteral vaccines are observed to be less effective in the developing world, the true impact of malnutrition, environmental enteropathy (EE), breast feeding, and coinfections is currently not well defined but a matter of ongoing clinical studies [111–116]. A currently proposed model is that altered condition of gut mucosa, microbiome , and metabolome negatively affects vaccine efficacy.

Protein CAMK4 (CaMKIV) may be part of the link between adaptive immunity, vaccines, and the nervous system. Fold change at day 3 post-vaccination with TIV is negatively correlated with antibody titers at day 28 via reduction of plasmablast expansion, and it is a well-known factor in T-cells and neuronal memory consolidation [12, 117]. The autonomous nervous system and the immune system cross-talk via the neuro-immune axis, misregulation of which is implicated in hypertension and cardiovascular disease [118]. Part of this regulation is that the brain can sense inflammatory cytokines and can modulate immune response s. The neurotransmitter acetylcholine can significantly attenuate the release of pro-inflammatory cytokines [119]. The gut as an organ of both central immunological and metabolic function is also a critical link to the microbiome and the gut-brain axis is implicated in the development of autoimmune and neurodevelopmental disorders [120]. In mice it has also been shown that defects in TLR5 lead to an altered microbiome and so-called metabolic syndrome (MeS) and transplantation of this altered microbiome to wild-type mice also confer features of MeS [121]. It is also becoming clear that there is a significant feedback between the endocrine system and the microbiome, impacting on metabolism and immunity [122]. To further stress the role of the gut and signaling molecules aside of proteins in immune homeostasis, the role of bile acids as metabolic regulators with a part in inflammatory disorders and relationship to the gut microbiome has recently been reviewed [123]. In fact there is substantial evidence that psychological distress can predict gastrointestinal disorders and vice versa; animal studies suggest the role of early life conditioning in later health and disease. In this context it is important to better understand potential effects early life vaccination plans may have on shaping microbiota and hence this may lead to improved timing of vaccination schedules [124]. Major depressive disorders have been associated with the so-called metabolic syndrome and low-grade inflammation in the central nervous system including a role of adipokines leptin and ghrelin [125, 126]. The link between immunity and the nervous system is also of relevance because of the recurring assertions of vaccines contributing to certain neurological conditions. However, recently stronger vaccine responses against meningococcal conjugate vaccine in children with symptoms of depression and anxiety have been reported [127]. At this point it may be speculated whether the reported pro-inflammatory signature of carbohydrate-containing vaccines (MedImmune, Menactra) may be enhanced by a pre-existing low-grade inflammatory condition and which effect this would have on vaccines requiring different signatures such as TIV [2]. Another recent development is the potential recognition of narcolepsy as an autoimmune disorder. Narcolepsy has recently been associated with infectious diseases and specifically the 2009 H1N1 (swine flu) pandemic and associated vaccine. This link does not seem to be definite, however, and may depend on genetic and ethnical background including the HLA-DQB1*06:02 genotype, additional unknown cofactors, and possibly vaccine formulation [128–131].

Taken together, metabolism and nervous and immune system are tightly interacting units. In fact, differentiating them may be more the result of an artificial concept than of a biological reality. Considering the known role of some hormones and metabolic disorders on vaccine efficacy this suggests that input from beyond classical immune cells may significantly contribute to the design of future systems vaccinology studies. Sampled tissues and data types will have to reflect the system a vaccine has to operate on. Specifically metabolomics, proteomics, and lipidomics should be valuable additions to currently pursued procedures in systems vaccinology. Yet it is currently unclear whether systemic determinants such as accessible in the blood-serum metabolome and lipidome may be sufficient to predict implications on vaccine responses, or whether local distribution is critical and well predicted by systemic concentration.

6 Glycans and Immunity

Recently the role of glycans in the immune system has been reviewed by Maverakis and colleagues, pointing out the various critical implications the glycocalyx of eukaryotic cells as well as glycosylations of serum proteins including antibodies have on modulating immunity [132]. While few data are available, analyzed in relationship to vaccines the implications for immunity have become clear, signaling the need to include the glycome in future systems vaccinology analysis. Indeed the glycome is heavily underrepresented in the majority of current systems biology investigations, which is arguably caused by experimental complexity. Yet current evidence shows that it is of critical importance in modulation of immunity and may provide numerous markers of use in understanding current and future reaction of individuals to vaccines. Also it should be considered that each antigen may be target of numerous antibodies where each may potentially be differently glycosylated, potentially grossly altering the effect from pro- to anti-inflammatory or vice versa. High-dose intravenous immunoglobulin (IVIg) therapy is used to treat autoimmune disorders and transplant rejection where the effect is assumed to rest on anti-inflammatory antibodies, or specifically IgG with preference of anti-inflammatory Fc receptors [133]. The generation of anti-inflammatory glycosylation of IgG (specifically with terminal sialic acid) in tolerogenic therapies has recently been demonstrated to rest on antibody development in a non-inflammatory environment, suggesting the use of systems vaccinology for the deeper understanding of involved mechanisms and development of potentially supportive anti-inflammatory adjuvants [134]. Although the entire glycome can be analyzed using mass spectrometric approaches and great technical advances have recently been made using lectin microarrays and capillary electrophoresis, analysis of position-specific glycosylations is hindered by the non-template-based nature of glycosylations, as Maverakis et al. point out [135, 136]. Now we are in the unsatisfying situation to know there is a critical component to understanding immunity, but essentially lack tools equivalent in ease of use to other omics technologies. Optimal resolution of antibody class, isotype, and relative abundance of these should be part of any comprehensive analysis of antibody-mediated immune reactions. High titers do not necessarily mean desired effect if Fc regions of generated antibodies do not activate the intended lectins and/or Fc receptors [137, 138]. It is known that immunoglobulin Fc regions can take on hundreds of different structures with slightly or gravely different effects on targeted cell types and hence achieved effect in cancer, autoimmunity, and infectious diseases. Profiling of adjuvants should therefore consider the precise nature of produced antibodies, as biomedical effects can be diverse and at least theoretically inverse to the intended. Changes in immunoglobulin glycosylation have been described for numerous autoimmune disorders and infectious diseases [139–141]. While there has to the author’s knowledge been no definite proof that these changes are causative of disease they reflect changes in the immune system pinpointing towards biomarkers and very possibly at least contribute to development of an unbalanced immune phenotype. This assumption is based on the well-established dependence of immunoglobulin affinity to Fc receptors based on subclass and Fc glycosylation pattern as well as glycosylation of the receptor and resulting modification of antibody effect on various immune cell types [142, 143]. At least in humans pro- and anti-inflammatory effect of Fc gammaRII receptor isoforms (FcγRIIa and FcγRIIb, respectively) is well established [137]. In addition these receptors are differently responsive to single antibodies, where the majority is only responsive to immune complexes [132]. The overall effect of a particular antibody should therefore depend on affinity to Fc receptors (particularly pro- and anti-inflammatory) and relative abundance of these Fc receptors on specific target cells. Therefore unless an antibody interacts solely with FcγRIIb it may still also elicit pro-inflammatory signals.

In the case of the human pathogen dengue virus where antibody-dependent enhancement (ADE) is currently understood as a major driver of pathology, the role of FcγRIIA may primarily be enrichment of virus/antibody complexes on the cell surface [42]. However the second receptor variant FcγRIIb (that stimulates an anti-inflammatory effect) has been suggested to provide only limited ADE effect in spite of equivalent antibody Fc affinity, implicating that subtype of generated antibodies in dengue natural infection and likely also dengue vaccines may critically impact pathology and efficacy of the vaccines [144].

7 Conclusions

The last two decades have brought tremendous advancements in omics technology. Next-generation sequencing (NGS) but also microarray-based transcriptomics and peptide chips, proteomics, lipidomics, and glycomics have either been invented or significantly matured. Since the emergence of these technologies resulted in tremendous data volumes which need to be stored, analyzed, and brought into context the new science of system biology has emerged. The basic idea is to avoid reductionist approaches and view the entire investigated system and all measurable components in parallel to effectively observe system perturbations and cross talk. At the same time these advances have made immunology leap ahead, now allowing seeing many processes of health and disease in the context of immunity. In many ways related arts and in this context especially vaccine development have sought to close the gap between basic research and application. While this is not always easy we now see the emergence of another fusion of sciences: systems vaccinology. Combining the methodologies of systems biology with the tools for interpretation stemming from modern immunology and focusing specifically at the improved understanding and development of vaccines this science has the potential to be a game changer for vaccine development. Many areas which have so far suffered from the immense complexity, interdependence and dynamic interplay of gene expression, interacting cell and tissue types, secretomes, metabolome, lipidome, genetic background, and immunological history present in immune systems are now offered an integrative approach to understand success and failure of vaccines and parameters to predict these early on. As such systems vaccinology has the intention to enhance understanding of vaccines, improve development processes of new vaccines, and possibly support clinical trials and personalized medicine. Specifically early identification of nonresponders and patient stratification in clinical trials may reduce the risk of future late-stage vaccine failures or should at least allow understanding of the reasons for failure. Also in already licensed vaccines systems vaccinology will allow deeper insights into determinants of efficacy including metabolic disorders especially those related to obesity, microbiome , and malnutrition. Ultimately it is a highly integrative development which will attempt to include as many views and technological options as possible, crossing barriers between disciplines. In fact modern immunology is already doing that, specifically when linking low-grade inflammation to metabolism and immune competence. This new understanding of disorders such as cardiovascular disease, type II diabetes and gastrointestinal dysfunction and their relationship with the microbiome is a major highlight of modern immunology. Therefore systems vaccinology only has to follow suit to bring modern vaccinology into a new era by integrating the impact of genes, environment, and the microbiome on protective immunity induced by vaccination.