Emerging technologies for systems vaccinology – multi-omics integration and single-cell (epi)genomic profiling

Abstract

Systems vaccinology leverages high-throughput “omics” technologies, such as transcriptomics, metabolomics, and mass cytometry, coupled with computational approaches to construct a global map of the complex processes that occur during an immune response to vaccination. Its goal is to define the mechanisms of protective immunity and to identify cellular and molecular correlates of vaccine efficacy. Emerging technological advances including integration of multi-omics datasets, and single-cell genomic and epigenomic profiling of immune responses, have invigorated systems vaccinology, and provide new insights into the mechanisms by which the cellular and molecular information underlying immune memory is stored in the innate and adaptive immune systems. Here, we will review these emerging directions in systems vaccinology, with a particular focus on the epigenome, and its impact on modulating vaccination induced memory in the innate and adaptive immune systems.

Introduction

Vaccination is one of the most cost-effective public health tools in medicine, and has eradicated smallpox, and saved countless lives from other infectious diseases. Yet, despite these past successes, vaccines against major global health threats, such as HIV, malaria, or influenza, are lacking. A major challenge in the design of novel vaccines is our limited understanding of the complex molecular and cellular processes that determine lasting vaccine protection. Traditionally, immunologists have approached this problem with a reductionist approach, purifying certain immune cell types or molecules and studying them in isolation. Recent efforts, however, have highlighted a systems biological approach, which focuses on studying interactions between all components of a complex system to gain a holistic view of the immune response. The field of systems vaccinology applies these ideas to the immune system with two aims: 1) to identify molecular or cellular signatures that predict vaccine responses and 2) to delineate the molecular and cellular mechanisms that generate durable protective immunity. Aim 1) has the potential to profoundly reduce the time and costs of clinical trials assessing vaccine efficacy. Currently, efficacy is assessed in large phase 3 clinical trials involving thousands of subjects who receive the vaccine or placebo, and are then followed for several years in order to measure the incidence of disease. Instead, if it were possible to identify an early signature or biomarker that was induced a few days after vaccination, that could predict the efficacy of a vaccine, this would greatly accelerate the process of vaccine testing and development. Aim 2) strives to provide novel mechanistic insights into how vaccines induce protective immunity. These insights can then guide novel strategies for vaccine design.

Systems vaccinology studies typically leverage high-throughput technologies that allow the quantitative measurement of many molecules or cell types - such as transcriptomics, multiplexed cytokine measurement, or multiparameter flow or mass cytometry - in an attempt to describe the system as completely as possible. Researchers then apply computational and mathematical approaches to construct models of the interactions between various immune system components. When applied to cohorts of vaccinated participants or animals, systems vaccinologists use these models to determine correlates of protection and to create and experimentally test hypotheses of the molecular and cellular events underlying protective immunity.

Recently, there has been a growing interest in establishing the epigenetic mechanisms that impact immune responses to vaccines. The epigenome is a central orchestrator of gene expression, and emerging evidence suggests that it has a significant impact on the functional activity of immune cells. Furthermore, immune cells can maintain epigenetic modifications over prolonged periods, making this phenomenon extremely interesting in the context of vaccination, which aims to induce durable changes to the immune system. Here, we give a brief overview of the nascent field of systems vaccinology and discuss novel concepts and technological advances.

A brief history of systems vaccinology

The first studies that applied systems-based approaches to investigate vaccine-induced immune responses used the yellow fever vaccine YF-17D [1,2]. YF-17D is a live attenuated viral vaccine and one of the most effective vaccines ever developed [3]. It induces protection in >99% of individuals and generates neutralizing antibody responses that can persist up to 30 years [3]. Despite its efficacy, the precise molecular and cellular mechanisms by which YF-17D stimulates immune responses were poorly understood. We previously demonstrated that YF-17D activates multiple subsets of dendritic cells by signaling via Toll like receptors 2, 3, 7, 8, and 9, and that signaling via TLRs was essential for its immunogenicity [4]. In 2008, we applied omics technologies using transcriptomics, multiplex analysis of serum cytokines, and multi-dimensional flow cytometry, to profile the innate and adaptive responses to vaccination in the blood of healthy humans vaccinated with YF-17D [1]. The results demonstrated that YF-17D induced a high magnitude antigen-specific CD8+ T cell response, as well as robust neutralization antibody titers. Furthermore, as might be expected from a live attenuated viral vaccine, YF-17D induced a strong antiviral gene signature, including type I IFN response, within 3 to 7 days of vaccination. Using computational approaches, we identified transcriptional signatures induced early within a few days of vaccination, that correlated with the magnitude of the ensuing antigen specific CD8+ T cell response and neutralizing antibody response. Furthermore, the capacity of these signatures to predict the magnitude of the antigen-specific CD8+ T cell response and neutralizing antibody response was measured using machine learning approaches in an independent study of subjects who received the yellow fever vaccine. These results provide proof of concept that systems biological approaches can be used to identify molecular signatures that predict vaccine immunogenicity. In an independent study using a comparable approach, Gaucher et al identified similar signatures including complement factors and interferon pathways [2].

Following the studies with YF-17D, we and several other labs used systems vaccinology approaches in short succession to investigate immune responses to vaccination against various diseases including smallpox [5], malaria [6–8], HIV [9], Meningococcus [10], Pneumococcus [11], and varicella-zoster [12,13]. Notably, several studies successfully combined systems-based approaches with human challenge models to define novel correlates of protective vaccine efficacy [6–8]. A particular focus in these efforts was put on the seasonal influenza vaccine, which is frequently administered but varies considerably in its effectiveness from year to year [14]. Applying systems approaches to the influenza vaccine, we and others identified early transcriptional signatures that correlated with specific antibody responses and served as starting points for subsequent research into approaches to improve vaccine efficacy [11,15–19].

A significant conceptual advance in recent years was the shift from a single gene-focused analysis to a pathway- and gene module-centered analysis. In this approach, correlation analysis identifies genes that behave similarly throughout vaccination to build gene interaction networks. These gene interaction networks are then correlated with immunological and clinical measurements to extract biological insights. In comparison to using individual genes as correlates, this network-centered approach captures global, biological processes that are more robust to biological and technical noise [10,20]. Due to the reduced noise, gene network-based analysis (termed “Blood Transcriptional Modules”- or BTM-based analysis in one study [10]) enables the comparison of vaccine cohorts from different studies and labs. Using the BTM approach, we set out to determine whether a universal gene signature that could predict protective responses for any given vaccine existed. We collected gene expression data from five different vaccine cohorts (polysaccharide and conjugated meningococcus vaccines, yellow fever vaccine, inactivated seasonal influenza vaccine, and live attenuated influenza vaccine), and calculated the activity score for all BTMs, in each vaccine. We then correlated the activity scores with vaccine-specific antibody responses to identify BTMs associated with vaccine efficacy. With the distinct nature of these vaccines and variations between cohorts, one might have expected a unique antibody correlation profile for each vaccine, as suggested by earlier pathway analyses without BTMs [10]. Instead, we identified three distinctive signatures of early BTMs that correlated with antibody titers; surprisingly, some of these were shared by multiple vaccines. Further analysis revealed that vaccines, which share specific immunogenic characteristics, such as the polysaccharide and con jugated meningococcus vaccine, activate similar biological pathways to induce antibody production. Distinct vaccines, such as YF-17D, in contrast, seem to employ unique pathways to induce protection. These results demonstrate that BTM analysis can identify shared vaccine signature and could become an important tool in the early evaluation of novel vaccines.

Finally, systems vaccinology approaches are also yielding new biological insights about the human immune system. For instance, identification of transcriptional correlates of humoral immunity to the seasonal influenza vaccine (TIV) revealed that early expression of TLR5, an innate pathogen receptor specific for bacterial flagellin, was highly associated with the day 30 antibody response [15]. This was surprising as TIV is an inactivated viral vaccine lacking flagellin. Mechanistic follow-up studies in gene knockout mice demonstrated that TLR5 mediated innate sensing of bacterial flagellin in the gut microbiome provided an adjuvant signal that enhanced the antibody response [21]. To determine if the gut microbiome was also crucial in modulating antibody responses in humans, we conducted a clinical trial in which healthy human subjects received a cocktail of antibiotics for five days [22]. We then administered TIV on the fourth day of antibiotics treatment. Antibiotics administration resulted in a transient, 10,000-fold reduction in the total quantity of gut bacteria, and long-lasting changes in bacterial diversity. Importantly, there was a marked reduction in the vaccine-specific IgG and IgA antibody titers -but only in subjects with low baseline antibody titers against influenza and no exposure to influenza vaccine or infection during the three preceding flu seasons. These results suggest that immunological imprinting, caused by prior exposure to influenza vaccination or infection, can withstand even the most severe gut dysbiosis [23]. In addition to this effect on the adaptive immune system, antibiotics resulted in enhanced inflammatory signatures of innate immunity [22], similar to that observed in elderly subjects [19]. Strikingly, antibiotics also resulted in a profound reduction in secondary bile acids, and the concentration of secondary bile acids correlated inversely with the inflammatory signatures [22]. Consistent with other reports, these findings position secondary bile acids as potential key regulators of inflammation and inflammasome activation [24].

Recent technological advances

Integrative analysis of multi-omics data

While most early systems vaccinology studies used gene expression and multiparameter flow cytometry data to investigate immune responses, more recent studies have integrated additional data types, including genomic variants [16], plasma metabolites [12,22], and the microbiome [22], as appropriate technologies became available. In the context of these “multi”-omics approaches, an essential advancement was the development of multifactorial computational tools that enabled the integration of orthogonal datasets to study interactions between different biological structures and their collective impact on vaccine efficacy [12,22].

Using such a multifactorial model, our lab, for the first time, combined data from transcriptomics, metabolomics, and immunological assays to investigate the integrated immune response to the varicella-zoster vaccine (VZV). At that time, it was unclear how metabolic processes would affect vaccination. Surprisingly, we discovered that the transcriptional and metabolic responses towards VZV are tightly coordinated and highly correlated [12]. From our analysis, a network of metabolites and genes involved in cholesterol biosynthesis emerged that integrated and predicted the humoral and cellular immunity to VZV. Similarly, in our analysis of the effect of antibiotics on vaccine responses, multifactorial models revealed an integrated network of gut bacteria, metabolites, and gene expression with impact on inflammation [22]. These results suggest that the metabolome and the microbiota can act as critical regulators of immune responses and position multifactorial analyses as an efficient tool for systems vaccinology. Future studies have to reveal how vaccines could harness these effects for improved protection.

Single-cell profiling

Recent technological advances extend the immunologist’s single-cell toolbox for genomics assays that allow researchers, for the first time, to interrogate entire molecule classes (e.g., transcriptomes, genomes, chromatin states) at the single-cell level. Several groups used single-cell RNA-seq to construct detailed single-cell maps of the innate immune cell compartment in healthy individuals [25,26]. When clustering and annotating the detected cells, the authors discovered a novel subset of dendritic cells that displayed progenitor capacities and could potentially be targeted to improve cancer vaccination strategies. Subsequently, this subset was further described and validated on the protein level via mass cytometry [26,27]. Using similar approaches to investigate the molecular events controlling early immune activation, we and others constructed single-cell maps of transcriptome of in vitro stimulated innate immune cells [28,29]. We discovered that a complex paracrine signaling network controls the magnitude of the interferon response, which could potentially inform future vaccine designs aimed at inducing antiviral or anticancer signatures. Other efforts focused on exploring the single-cell landscape of the adaptive immune system transcriptome in the context of vaccines. Constructing maps of purified T cell populations from vaccinated participants, for instance, researchers identified previously unknown memory subsets, some of which were linked to extended vaccine protection [30,31]. Lau et al. applied a similar strategy to circulating B cells and identified a subset of B cells that could be predisposed to differentiate into long-lived plasma cells [32].

While substantially contributing to our understanding of basic immunological processes at the single-cell level, these studies each focused on one particular immune cell type only, often in an in vitro setting. A comprehensive systems-based survey, describing the entire immune system and its interactions at the single-cell level and in the context of in vivo vaccine responses, is currently lacking. For such a study, it would be essential to address questions regarding 1) the impact of cellular heterogeneity on vaccine efficacy, 2) the role of previously unappreciated immune cell subsets in the context of vaccine responses, and 3) the interconnection of transcriptional and regulatory programs in different cell types. Recently, a first study set out to answer such questions in the context of natural HIV infection and reported on previously unknown subsets of cycling NK cells and polyfunctional monocytes in individuals who developed spontaneous viral control [33].

Epigenetic regulation of innate and adaptive immune memory

Genomic variations, in the form of single nucleotide polymorphisms, have previously been shown to impact the immune response to influenza vaccination [16]. To determine the extent of which these genetic variations impact immune responses to vaccination, Brodin et al. analyzed mono- and dizygotic twins and determined various immunological factors, including the production of antibodies in response to seasonal influenza vaccination [34]. Their analysis revealed that non-heritable factors dominated almost 80% of the observed variability. Furthermore, the degree of variability between twins increased with age, suggesting that an accumulation of environmental influences over time might be a driving factor. Recent attention has focused on the epigenome as a critical hub of non-genetic and environmentally induced variability in the immune system. Cheung et al., for instance, demonstrated in a twin cohort that the chromatin landscape of blood circulating immune cells differs considerably between individuals and that this variability is mostly due to non-heritable factors [35]. Similar to Brodin, Cheung observed an increase in epigenetic variability with age, indicating an environmental impact. Further studies showed that many environmental factors, such as pathogens [36], pollutants, nutrition, and other stressors, can modify the epigenome [37]. In light of these findings, an important question is to what extent epigenetic changes affect human immune responses to vaccines.

The epigenome consists of a set of chemical modifications that are attached to the chromatin (DNA + histone proteins) [38]. Their type and position regulate chromatin structure and accessibility as well as the recruitment of proteins such as activators, transcription factors or polymerases. These processes have a substantial impact on gene activity, making the epigenome a central hub for gene regulation. Epigenomic events have been shown to impact many fundamental biological processes, including development, differentiation, and cancer [38,39]. Importantly, the epigenome can store information for prolonged periods and over generations of cells (and possibly organisms) [38]. In the context of vaccines, which aim to induce lasting immune protection, these features are of particular interest.

Currently, there are four major ways to assess the epigenomic landscape of a cell: measuring the levels and positions of chemical attachments to 1) histone molecules (ChIP-seq, EpiTOF) (Figure 1a), or 2) DNA (e.g., Bisulfite sequencing) (Figure 1b), 3) measuring chromatin structure (open = active, closed = inactive, e.g., DNAse-seq, FAIRE-seq, ATAC-seq) (Figure 1c), and 4) measuring the 3D genome organization (determination of regulatory interactions between distant genomic regions, e.g., HiC) (Figure 1d). A full review of these techniques would exceed the scope of this manuscript, and we refer the reader to several excellent texts for detailed information [35,38–40]. Of the mentioned techniques, ChiP-seq, ATAC-seq, and EpiTOF are amongst the most frequently used for immunological analyses. They are high-throughput (EpiTOF: ~ 20 samples of 2million cells each in one experiment), require relatively few input cells (ATAC-seq: only ~5000 cells), or enable high content analyses (ChIP-seq: measuring histone modifications genome-wide). In the context of vaccines, Mihai Netea’s lab recently showed that the tuberculosis vaccine BCG induces changes to the epigenomic landscape of monocytes in humans and mice [41,42]. Using ChIP-seq, they found that histones at several proinflammatory gene loci showed increased levels of H3K4me3 – a histone mark indicative for poised or active promoters [41]. Importantly, these marks were still modified three months after vaccination, suggesting that the observed changes were long-lasting. In a follow-up study, Arts et al. demonstrated that the same vaccine also induced a net increase in H3K27ac – a histone mark indicative for active enhancers and gene transcription [42]. Both studies found that histone changes coincide with increased production of proinflammatory cytokines upon ex-vivo re-stimulation. When injecting participants with YF-17D, a live-attenuated vaccine, 30 days after BCG vaccination, Arts et al. observed a modest decrease in viral titers. This result indicates that vaccine-induced epigenetic reprogramming might convey non-specific protection, as previously observed in mouse and in vitro studies [41,43]. Intriguingly, in this context, epidemiological data were interpreted to suggest that several vaccines, including BCG and measles vaccines, convey broad protection against unrelated pathogens [44]. Based on these results, Mihai Netea and colleagues recently hypothesized that not only the adaptive but also the innate immune system can build immunological memory – based on epigenetic reprogramming [45]. The significance of innate immune memory for global health and vaccine design is yet to be determined and there are still many questions around the biological factors that come into play in mediating these effects. It is, for instance, not well understood how monocytes, which have a lifetime of 1–7d, can maintain epigenetic changes for more than three months. Also, it is unclear which vaccines can induce epigenetic changes and whether it is possible to design compounds that deliberately and specifically induce favorable epigenetic profiles.

Figure 1 –

Epigenetic mechanisms of gene regulation and how to measure them.

Another focus of epigenetic analysis in the context of vaccines has been the immunological events leading to the generation of long-lived T and B cell memory. A fundamental question in the field is the origin of long-lived CD8+ memory T cells. While crucial for the maintenance of protection against viral infections, the precise mechanisms that lead to the generation of these cells are currently unknown, hampering the rational design of vaccines. Two studies from Rafi Ahmed’s lab combined Bisulfite sequencing and ATAC-seq to study the epigenome of antigen-specific T cells after vaccination with YF-17D, which is known to induce potent, long-lasting T cell responses [46,47]. The authors found that long-lived memory T cells, while phenotypically similar to naïve T cells, maintain an epigenomic profile closely resembling that of effector T cells generated during the acute vaccine response. Based on these findings, they conclude that long-lived memory T cells differentiate from effector T cells during or after the infection and do not arise via a parallel pathway from naïve T cells. Intriguingly, the observed epigenetic states were stable for more than a decade. Several other studies emerged that used single-cell ATAC-seq to highlight the importance of epigenetic modifications for the induction and maintenance of T cell differentiation states [48–50].

While these studies contributed significantly to our understanding of the epigenomic mechanisms that regulate specific aspects of immunity, a comprehensive systems-based survey of the chromatin landscape during vaccination is currently missing. With suitable technology finally available, systems vaccinologists can now probe the chromatin state of many different immune cell subsets in parallel and throughout the entire course of vaccination. The challenge will then be to interpret this wealth of epigenomic data and, most importantly, to understand how it affects clinical aspects such as vaccine protection.

Conclusion

Systems vaccinology has proven to be a promising tool in determining novel correlates and mechanisms of vaccine protection. Rapid technological advances are facilitating the interrogation of the immune system in exquisite detail, and at all levels of biological regulation. In this context, there has been heightened interest in the epigenome as a central regulatory hub within immune cells. These outlooks promise exciting biological discoveries, but the success of the field will eventually need to be measured by the translation of its findings into novel vaccine approaches. Would it be possible, for example, to administer small molecule adjuvants that modulate the epigenetic profile of myeloid cells (“Epijuvants”)? Can we use systems vaccinology approaches to answer these questions more efficiently and cost-effective? One thing is sure: it is an exciting time to be a systems vaccinologist.

Highlights.

1. Systems vaccinology offers a unique ways to study immunity to vaccination.

2. New technologies such as single-cell profiling are invigorating systems vaccinology.

3. Such technologies are revealing mechanisms of epigenetic control of innate memory.