Selecting an optimal antigen is a crucial step in vaccine development, significantly influencing both the vaccine’s effectiveness and the breadth of protection it provides. High antigen sequence variability, as seen in pathogens like rhinovirus, HIV, influenza virus, complicates the design of a single cross-protective antigen. Consequently, vaccination with a single antigen molecule often confers protection against only a single variant. In this study, machine learning methods were applied to the design of factor H binding protein (fHbp), an antigen from the bacterial pathogen Neisseria meningitidis. The vast number of potential antigen mutants presents a significant challenge for improving fHbp antigenicity. Moreover, limited data on antigen-antibody binding in public databases constrains the training of machine learning models. To address these challenges, we used computational models to predict fHbp properties and machine learning was applied to select both the most promising and informative mutants using a Gaussian process (GP) model. These mutants were experimentally evaluated to both confirm promising leads and refine the machine learning model for future iterations. In our current model, mutants were designed that enabled the transfer of fHbp v1.1 specific conformational epitopes onto fHbp v3.28, while maintaining binding to overlapping cross-reactive epitopes. The top mutant identified underwent biophysical and x-ray crystallographic characterization to confirm that the overall structure of fHbp was maintained throughout this epitope engineering experiment. The integrated strategy presented here could form the basis of a next-generation, iterative antigen design platform, potentially accelerating the development of new broadly protective vaccines.

Background Determinants of maternal-fetal cytomegalovirus (CMV) transmission and factors influencing the severity of congenital CMV (cCMV) infection are not well understood.Methods We conducted a descriptive, multicenter study in pregnant women >= 18 years old with primary CMV infection and their newborns to explore maternal immune responses to CMV and determine potential immunologic/virologic correlates of cCMV following primary infection during pregnancy. We developed alternative approaches looking into univariate/multivariate factors associated with cCMV, including a participant clustering/stratification approach and an interpretable predictive model-based approach using trained decision trees for risk prediction (post hoc analyses).Results Pregnant women were grouped in 3 distinct clusters with similar baseline characteristics, particularly gestational age at diagnosis. We observed a trend for higher viral loads in urine and saliva samples from mothers of infants with cCMV versus without cCMV. When using a trained predictive-model approach that accounts for interaction effects between variables, anti-pentamer immunoglobulin G antibody concentration and viral load in saliva were identified as biomarkers jointly associated with the risk of maternal-fetal CMV transmission.Conclusions We identified biomarkers of CMV maternal-fetal transmission. After validation in larger studies, our findings will guide the management of primary infection during pregnancy and the development of vaccines against cCMV.Clinical Trials Registration NCT01251744. Using different models (participant clustering/stratification and an interpretable predictive model using trained decision trees for risk prediction), we found that anti-pentamer IgG concentration during pregnancy and viral load in saliva are biomarkers jointly predicting the risk of maternal-fetal cytomegalovirus transmission. Graphical Abstract

Vaccines are essential tools in the control of infectious diseases. The development of vaccines has matured from a mostly empirical to a highly sophisticated approach that is science-grounded and based on state-of-the-art knowledge in molecular biology, immunology, and structural biology. Insights in the mechanism of action of vaccines have been driven by clinical research studies that leverage systems vaccinology approaches to unravel the links between the innate immune response and the quality and persistence of adaptive immunity. Over the last decade an increasing number of clinical systems vaccinology studies have been published. In parallel, developments in artificial intelligence and machine learning (AI/ML) have made a massive impact in real-life applications from image classification to speech recognition. AI/ML has also entered the world of vaccinology. In this manuscript, we reflect on how AI/ML could be used to leverage the wealth of clinical systems biology data to drive novel insights in vaccines mechanism of action.

Transcriptional responses to adjuvanted vaccines can vary substantially among populations. Interindividual diversity in levels of pathogen exposure, and thus of cell-mediated immunological memory at baseline, may be an important determinant of population differences in vaccine responses. Adjuvant System AS01 is used in licensed or candidate vaccines for several diseases and populations, yet the impact of pre-existing immunity on its adjuvanticity remains to be elucidated. In this exploratory post-hoc analysis of clinical trial samples (clinicalTrials.gov: NCT01424501), we compared gene expression patterns elicited by two immunizations with the candidate tuberculosis (TB) vaccine M72/AS01, between three groups of individuals with different levels of memory responses to TB antigens before vaccination. Analyzed were one group of TB-disease-treated individuals, and two groups of TB-disease-naïve individuals who were (based on purified protein derivative [PPD] skin-test results) stratified into PPD-positive and PPD-negative groups. Although TB-disease-treated individuals displayed slightly stronger transcriptional responses after each vaccine dose, functional gene signatures were overall not distinctly different between groups. Considering the similarities with the signatures found previously for other AS01-adjuvanted vaccines, many features of the response appeared to be adjuvant-driven. Across groups, cell proliferation-related signals at 7 days post-dose 1 were associated with increased anti-M72 antibody response magnitudes. These early signals were stronger in the TB-disease-treated group as compared to both TB-disease-naïve groups. Interindividual homogeneity in gene expression levels was also higher for TB-disease-treated individuals post-dose 1, but increased in all groups post-dose 2 to attain similar levels between the three groups. Altogether, strong cell-mediated memory responses at baseline accelerated and amplified transcriptional responses to a single dose of this AS01-adjuvanted vaccine, resulting in more homogenous gene expression levels among the highly-primed individuals as compared to the disease-naïve individuals. However, after a second vaccination, response heterogeneity decreased and was similar across groups, irrespective of the degree of immune memory acquired at baseline. This information can support the design and analysis of future clinical trials evaluating AS01-adjuvanted vaccines.

Replication-deficient chimpanzee adenovirus (ChAd) vectors represent an attractive vaccine platform and are thus employed as vaccine candidates against several infectious diseases. Since inducing effective immunity depends on the interplay between innate and adaptive immunity, a deeper understanding of innate immune responses elicited by intramuscularly injected ChAd vectors in tissues can advance the platform’s development. Using different candidate vaccines based on the Group C ChAd type 155 (ChAd155) vector, we characterized early immune responses in injected muscles and draining lymph nodes (dLNs) from mice, and complemented these analyses by evaluating cytokine responses and gene expression patterns in peripheral blood from ChAd155-injected macaques. In mice, vector DNA levels gradually decreased post-immunization, but local transgene mRNA expression exhibited two transient peaks [at 6 h and Day (D)5], which were most obvious in dLNs. This dynamic pattern was mirrored by the innate responses in tissues, which developed as early as 1–3 h (cytokines/chemokines) or D1 (immune cells) post-vaccination. They were characterized by a CCL2- and CXCL9/10-dominated chemokine profile, peaking at 6 h (with CXCL10/CCL2 signals also detectable in serum) and D7, and clear immune-cell infiltration peaks at D1/D2 and D6/D7. Experiments with a green fluorescent protein-expressing ChAd155 vector revealed infiltrating hematopoietic cell subsets at the injection site. Cell infiltrates comprised mostly monocytes in muscles, and NK cells, T cells, dendritic cells, monocytes, and B cells in dLNs. Similar bimodal dynamics were observed in whole-blood gene signatures in macaques: most of the 17 enriched immune/innate signaling pathways were significantly upregulated at D1 and D7 and downregulated at D3, and clustering analysis revealed stronger similarities between D1 and D7 signatures versus the D3 signature. Serum cytokine responses (CXCL10, IL1Ra, and low-level IFN-α) in macaques were predominantly observed at D1. Altogether, the early immune responses exhibited bimodal kinetics with transient peaks at D1/D2 and D6/D7, mostly with an IFN-associated signature, and these features were remarkably consistent across most analyzed parameters in murine tissues and macaque blood. These compelling observations reveal a novel aspect of the dynamics of innate immunity induced by ChAd155-vectored vaccines, and contribute to ongoing research to better understand how adenovectors can promote vaccine-induced immunity.

The RTS,S/AS01 vaccine provides partial protection against Plasmodium falciparum infection but determinants of protection and/or disease are unclear. Previously, anti-circumsporozoite protein (CSP) antibody titers and blood RNA signatures were associated with RTS,S/AS01 efficacy against controlled human malaria infection (CHMI). By analyzing host blood transcriptomes from five RTS,S vaccination CHMI studies, we demonstrate that the transcript ratio MX2/GPR183, measured 1 day after third immunization, discriminates protected from non-protected individuals. This ratiometric signature provides information that is complementary to anti-CSP titer levels for identifying RTS,S/AS01 immunized people who developed protective immunity and suggests a role for interferon and oxysterol signaling in the RTS,S mode of action.

The current routine use of adjuvants in human vaccines provides a strong incentive to increase our understanding of how adjuvants differ in their ability to stimulate innate immunity and consequently enhance vaccine immunogenicity. Here, we evaluated gene expression profiles in cells from whole blood elicited in naive subjects receiving the hepatitis B surface antigen formulated with different adjuvants. We identified a core innate gene signature emerging 1 day after the second vaccination and that was shared by the recipients of vaccines formulated with adjuvant systems AS01(B), AS01(E), or AS03. This core signature associated with the magnitude of the hepatitis B surface-specific antibody response and was characterized by positive regulation of genes associated with interferon-related responses or the innate cell compartment and by negative regulation of natural killer cell-associated genes. Analysis at the individual subject level revealed that the higher immunogenicity of AS01(B)-adjuvanted vaccine was linked to its ability to induce this signature in most vaccinees even after the first vaccination. Therefore, our data suggest that adjuvanticity is not strictly defined by the nature of the receptors or signaling pathways it activates but by the ability of the adjuvant to consistently induce a core inflammatory signature across individuals.

Background Vaccine clinical studies typically provide time-resolved data on adaptive response read-outs in response to the administration of that particular vaccine to a cohort of individuals. However, modeling such data is challenged by the properties of these time-resolved profiles such as non-linearity, scarcity of measurement points, scheduling of the vaccine at multiple time points. Linear Mixed Models (LMM) are often used for the analysis of longitudinal data but their use in these time-resolved immunological data is not common yet. Apart from the modeling challenges mentioned earlier, selection of the optimal model by using information-criterion-based measures is far from being straight-forward. The aim of this study is to provide guidelines for the application and selection of LMMs that deal with the challenging characteristics of the typical data sets in the field of vaccine clinical studies. Methods We used antibody measurements in response to Hepatitis-B vaccine with five different adjuvant formulations for demonstration purposes. We built piecewise-linear, piecewise-quadratic and cubic models with transformations of the axes with pre-selected or optimized knot locations where time is a numerical variable. We also investigated models where time is categorical and random effects are shared intercepts between different measurement points. We compared all models by using Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), Deviance Information Criterion (DIC), variations of conditional AIC and by visual inspection of the model fit in the light of prior biological information. Results There are various ways of dealing with the challenges of the data which have their own advantages and disadvantages. We explain these in detail here. Traditional information-criteria-based measures work well for the coarse selection of the model structure and complexity, however are not efficient at fine tuning of the complexity level of the random effects. Conclusions We show that common statistical measures for optimal model complexity are not sufficient. Rather, explicitly accounting for model purpose and biological interpretation is needed to arrive at relevant models. Trial Registration Clinical trial registration number for this study: NCT00805389, date of registration: December 9, 2008 (pro-active registration).

Article Figures and data Abstract eLife digest Introduction Results Discussion Materials and methods Data availability References Decision letter Author response Article and author information Metrics Abstract Systems vaccinology approaches have been used successfully to define early signatures of the vaccine-induced immune response. However, the possibility that transcriptomics can also identify a correlate or surrogate for vaccine inflammation has not been fully explored. We have compared four licensed vaccines with known safety profiles, as well as three agonists of Toll-like receptors (TLRs) with known inflammatory potential, to elucidate the transcriptomic profile of an acceptable response to vaccination versus that of an inflammatory reaction. In mice, we looked at the transcriptomic changes in muscle at the injection site, the lymph node that drained the muscle, and the peripheral blood mononuclear cells (PBMCs)isolated from the circulating blood from 4 hr after injection and over the next week. A detailed examination and comparative analysis of these transcriptomes revealed a set of novel biomarkers that are reflective of inflammation after vaccination. These biomarkers are readily measurable in the peripheral blood, providing useful surrogates of inflammation, and provide a way to select candidates with acceptable safety profiles. https://doi.org/10.7554/eLife.46149.001 eLife digest Measles, whooping cough and other diseases can cause serious illness and death in humans, especially in young children and other vulnerable individuals. Giving people vaccines ‘trains’ their immune system to recognize and fight the microbes that cause the conditions. During an infection, the immune system triggers a set of responses that limit the spread of the infectious agent and eliminate it from the body. This can include swelling of tissues (known as inflammation), which in rare cases, can be life threatening. Inoculations work by sparking a mild immune response in the body. Before a new vaccine is licensed for use, it is thoroughly tested in mice and rodents, and then in human volunteers, to ensure it will cause little or no inflammation. Finding a way to predict early on whether a vaccine candidate will trigger dangerous levels of inflammation would improve this process. To explore this, McKay, Cizmeci et al. injected the muscle tissue of different groups of mice with one of four licensed vaccines which, by definition, cause little or no inflammation. Other groups of animals were given one of three drugs known to trigger inflammation. Over the following seven days the team repeatedly collected blood as well as cells from the muscle tissue and the lymph nodes. These samples were then analysed to find out which genes were switched on or off at any given time. The experiments show that the responses of genes in the blood and lymph cells of the mice are connected to those in the muscle cells. Therefore, blood samples may provide a quick and convenient way to assess how an animal is responding to a potential new vaccine. By comparing the genes switched on or off in response to the different vaccines and drugs, McKay, Cizemeci et al. were able to identify a set of genes (known as “biomarkers”) that are associated with inflammation in animals. These biomarkers can be used to spot early on whether a new treatment is triggering inflammation. The next step would then be to identify a similar or identical set of biomarkers in other animals used in vaccine research, and in humans. Ultimately, this approach could make the assessment of the safety of a new vaccine candidate easier. https://doi.org/10.7554/eLife.46149.002 Introduction Systems biology approaches are increasingly being used to describe and define signatures of immunity, initially in the setting of infection but more recently in vaccine-induced responses, leading to the development of systems vaccinology (Querec et al., 2009; Olafsdottir et al., 2015). These transcriptomic analyses have primarily focused on the prediction of vaccine efficacy and immune outcome with the assessment of vaccine safety and potential reactogenicity relying on clinical reactive scores of adverse events (Bucasas et al., 2011; Furman et al., 2013; Obermoser et al., 2013; Vahey et al., 2010; Li et al., 2014; Leonardi et al., 2015; Vesikari et al., 2009). However, the transcriptome can also reveal the intricate details of the very early events after vaccination. We have utilized this approach to examine this initial innate response and any subsequent inflammation resulting from a vaccination, to identify biosignatures that have correlative or surrogate potential for vaccine-related inflammation and safety. It is likely that the activation and maturation of the fundamental responders within the immune system will follow set developmental patterns, which may be revealed by an examination of the transcriptome. Thus any potential differences between each vaccine would probably be due to the degree of response, and to the involvement of different populations of central immune system players and of accessory cells. The dominant determinants of such responses are the nature of the vaccine antigen, its formulation, and the presence or absence of molecularly defined adjuvants (Li et al., 2017; O'Gorman et al., 2015). The blood and lymphatic systems are the primary routes of transport of cells and chemical signals from one immunological organ to another, with immune cells and soluble mediators trafficking between the injection site and the local draining lymph node (Girard et al., 2012; Jackson, 2014). Indeed, proteins and molecules that can be measured in the blood after vaccination are likely to be useful biomarkers of such responses as blood draws can be readily taken from most individuals. Although a particular cell or inflammatory molecule may be present in relatively high numbers and concentrations in both the lymph node and the injection site, it is apparent that the large volume of blood in the circulatory system may dilute this molecule and prevent detection. In this study, we examined the molecular signatures generated in the blood, the draining lymph node and the injection site in the muscle of mice after the administration of a number of licensed vaccines, which set the benchmark for what is safe, and of a set of known agonists of Toll-like receptors (TLRs) or adjuvants, to set an upper boundary of inflammatory response (Haziot et al., 1998; Qureshi et al., 1999; Akira and Takeda, 2004; Fitzgerald et al., 2004; Matsumoto and Seya, 2008). The four licensed vaccines used in this study have a substantial history of safe use in humans, with large datasets supporting their safety and tolerability. They ranged from protein alone to vaccines that included adjuvants or a known component that have an acceptable degree of reactogenic potential (whole cell pertussis) (Frey et al., 2003; Tregnaghi et al., 2012; Mohanty et al., 2018; Van Den Ende et al., 2017). We also used the water-in-oil emulsion incomplete Freunds adjuvant (IFA) (Montanide ISA 51) and two TLR agonists (lipopolysaccharide (LPS) and Poly I:C) (Table 1). A set of indicators of vaccine-elicited inflammation in a small animal model will provide an early warning system that can be used to inform decisions on the potential tolerability and ultimately the utility of vaccines undergoing selection and development, and will provide additional tools to de-risk late-stage failure. Table 1 Vaccines and inflammatory agents. The vaccines, TLR agonists, adjuvant and saline used in the present study are detailed. Vaccines were administered by injection of 1/10th the human dose in a 50 µL volume. TLR agonists, IFA and saline were also given in a single 50 µL volume. All injections were into the mouse hind leg quadriceps muscle. https://doi.org/10.7554/eLife.46149.003 Vaccine/TLRComponentsAbbreviationManufacturerPentavac SDDiphtheria, tetanus, pertussis (whole cell), hepatitis B (rDNA) and haemophilus type b conjugate vaccinePentavacSerum Institute IndiaAgrippalTrivalent flu subunits – H3N2, H1N1 and influenza BTri-FluSeqirusFluadTrivalent flu subunits – H3N2, H1N1 and influenza B + MF59Tri-Flu + MF59SeqirusEngerix BRecombinant hepatitis B sAg absorbed on alumGlaxoSmithKlineIFAMontanide ISA 51 VGSeppicLPSLPS-EB UltrapureInvivogenPoly I:CPolyinosinic:polycytidylic acidSigma Results Transcriptomic profiles in muscle, draining lymph nodes and PBMCs of mice treated with different licensed vaccines and TLR agonists High-quality RNA samples, isolated from muscle tissue excised around the injection site, the draining medial iliac lymph nodes (MLN) and total peripheral blood mononuclear cells (PBMCs), were taken at 4, 8, 24, 48, 72, 168 hr after the vaccination or treatment with LPS, polyI:C or IFA (Table 1) and subjected to genome-wide transcriptome analysis. The unvaccinated mice, and mice receiving saline alone, were used as control groups. Differentially expressed genes were defined as those with a significant (Benjamini-Hochberg (BH) adjusted p-value<0.01) change in expression in a vaccinated group when compared to the unvaccinated group and the saline group at a given time point. Violin plots that chart the numbers of upregulated and downregulated differentially expressed genes revealed that the immunisations differed in terms of the magnitude and the kinetics of the transcriptomic responses induced in muscle, MLN, and blood (Figure 1). Strong responses were observed in the injected muscle tissue for all of the immunisations, with the highest levels of fold change and of absolute numbers of differentially expressed genes being detected at 72 and 168 hr after treatment. Pentavac SD, Poly I:C and particularly LPS also induced changes in differential gene expression in the muscle, MLN and blood at 4, 8, 24 and 48 hr post-injection. Engerix B and IFA elicited modest changes in the differential gene expression in the muscle at 48, 72 and 168 hr and at the later time points of 72 and 168 hr in the MLN, but very few differentially expressed genes were observed in the blood. Agrippal and Fluad exhibited similar patterns of differential gene expression at the later time points of 72 and 168 hr, but Fluad had an additional earlier signal at 8, 24 and 48 hr in the muscle (Figure 1). Figure 1 Download asset Open asset Overview of the transcriptomic responses to the injected vaccines and TLR agonists in the injected muscle, the draining lymph nodes (MLN), and the blood. The width of each violin is scaled to the number of genes that are differentially expressed at each given time point (BH-adjusted p-value <0.01). Differentially expressed genes were defined as those with a significant (BH-adjusted p-value <0.01) change in expression in a vaccinated group, when compared to the unvaccinated group and the saline group at a given time point. The upregulated or downregulated differentially expressed genes (Log2 fold-change) are presented for each tissue sampled and at each timepoint after injection of the vaccine or TLR agonist (n = 5 mice per timepoint). The vaccines or TLR agonists that were used are indicated on the right of the figure. IFA, incomplete Freunds adjuvant; LPS, lipopolysaccharide. https://doi.org/10.7554/eLife.46149.004 Distinctive transcriptional programs elicited after immunisation: gene-set enrichment and co-expression analysis Gene-set enrichment analysis was performed to identify the functional patterns of the biological processes that are induced by the different immunisations. Gene modules that are specific to the context of immune responses in blood tissue were previously defined by Li et al. (2014). These blood transcriptional modules were used to assess the signatures induced by the immunisations at the different time points. The blood transcriptional module enrichment profiles for every tissue are compared and presented in Figure 2 and Figure 2—figure supplements 1–3, where the gene sets with significant enrichment (lower than p<10−6) in each tissue are shown. In the injected muscle tissue, each vaccine or TLR agonist (with the exception of Agrippal) primarily elicited an upregulation of expression, as compared with the expresson levels seen after injection of the saline control, of genes related to inflammation, growth factors, innate immunity and cell damage (Figure 2—figure supplement 1). The induced transcriptomic response in the MLN was more refined, probably reflecting the effect of the migration of activated immune cells and local generalised stimulation by cyto/chemokine gradients. In these draining iliac lymph nodes, the injection of LPS, Pentavac SD and Poly I:C promoted an elevation of the interferon response genes as well as of cyto/chemokines (Figure 2—figure supplement 2). Finally, in the PBMCs, the transcriptomic analysis revealed that LPS, Pentavac SD and Poly I:C elicited an increase in gene transcripts that was associated with the interferon response and with activated monocytes and neutrophils (Figure 2—figure supplement 3). Figure 2 with 3 supplements see all Download asset Open asset Gene-set enrichment analysis reveals distinct mechanisms of transcriptional response to the injected vaccines and TLR agonists. Tissue data sets are presented as three coloured segments along the perimeter for the injected muscle (purple), the draining lymph nodes (green), and the blood compartment (red). An ordered list of blood transcriptional modules is labelled inside the outermost perimeter. Coloured bands adjacent to the module names correspond to the vaccines and Toll-like receptor (TLR) agonists (outside to inside bands): incomplete Freunds adjuvant (IFA) (grey), lipopolysaccharide (LPS) (blue), Pentavac SD (orange), Poly I:C (green), Fluad (red), Agrippal (purple), and Engerix (blue). For each band, there are six rings corresponding to time points 4, 8, 24, 48, 72, 168 hr (from outer to inner ring). The intensity of colours indicates the significance of the enrichment. Only modules with p-value for enrichment of <10−6 and an effect size (AUC) >0.8 are shown. DC, dendritic cell; ECM, extracellular matrix. https://doi.org/10.7554/eLife.46149.005 To identify transcriptional patterns in an unbiased way, gene modules were constructed using Weighted Gene Co-expression Network Analysis (WGCNA) (Langfelder and Horvath, 2008). Briefly, WGCNA discovers interacting genes by computing and matching the correlation patterns in their expression profiles over the entire time period, extracting clusters (modules) of highly correlated genes, and thereby categorizing the association between the modules and the traits (different vaccines). WGCNA was applied to the gene expression datasets for muscle, lymph nodes, and blood separately. The co-expression analyses identified 40 modules in muscle with numbers of genes ranging from 27 to 4979 (Figure 3—figure supplement 1A), 47 modules in lymph nodes with numbers of genes ranging from 26 to 3456 (Figure 3—figure supplement 1B), and 30 modules in blood with numbers of genes ranging from 24 to 9789 (Figure 3—figure supplement 1C). WGCNA assigns colours to name each module. We also added ‘mu’ for muscle, ‘ln’ for lymph nodes, or ‘bl’ for blood to each module name to avoid modules having the same name across different tissue datasets (such as muscle module (muM1_turquoise), lymph node module (lnM1_turquoise) and blood module (blM1_turquoise)). Figure 3—figure supplement 1 shows the correlations (r >0.3 positive correlations are in blue, r <−0.3 negative correlations are in red) between the modules and the immunisations. In muscle, co-expression analysis was able to identify modules that correlated with each immunisation, strongly (both positive and negative) for Pentavac SD and to a lesser degree for LPS, Fluad and Agrippal. These modules were different to those that correlated with a saline-only injection. In lymph nodes and blood, the method again revealed that Pentavac SD, LPS, Fluad and Agrippal treatments were associated with a number of positively and negatively significantly correlated modules. The connections between these highly correlated genes modules, which reveal whether the same sets of genes are co-regulated in the different tissues, are shown in Figure 3—figure supplement 1D, which identifies consensus modules associated with Agrippal, LPS and Pentavac SD. To interpret the extracted modules functionally, we compared the WGCNA modules with the reference blood transcriptional modules. A hypergeometric test was performed to quantify the overlap between the two sets of modules and the significant enrichments are reported in a heatmap (Figure 3). In muscle, the muM1-turquiose module contains genes that overlap with most of the reference modules and is associated positively with LPS and Pentavac SD; the muM2-blue module overlaps with reference modules of extracellular matrix (ECM), migration and mitochondria and is negatively associated with LPS and Pentavac SD (Figure 3 and Figure 3—figure supplement 1A). In lymph nodes, the lnM6-red module exhibits significant overlap with the reference modules associated with the cell cycle, whereas the lnM8-pink module is associated with interferon or antiviral sensing and this module contains genes that are co-regulated in LPS and Poly I:C immunisation (Figure 3 and Figure 3—figure supplement 1B). In blood, the blM1-turquoise module overlaps with reference modules of T cell, whereas the M4-yellow module overlaps with reference modules of monocytes, neutrophils and inflammatory/TLR/chemokines. The M1-turquoise and M4-yellow modules contain genes of similar transcriptional patterns that are significantly associated with LPS and Pentavac SD immunisation (Figure 3 and Figure 3—figure supplement 1C). Figure 3 with 1 supplement see all Download asset Open asset Heatmap representing the transcript overlap between WGCNA modules and the reference modules. Reference modules are the blood transcriptional modules defined by Li et al. (2014). These modules are shown in rows and are annotated within a higher functional group. Only WGCNA modules that showed a significant enrichment (hypergeometric test adjusted p-value<0.05) in any of the reference modules were included and shown in columns (hypergeometric test adjusted p-value<0.05). The legends on the figure report the strength of the p-value as a gradient of green. The colours at the top of the columnsindicate the tissue being analysed, whereas rows are colour-coded to indicate their higher annotation, as indicated in the key. https://doi.org/10.7554/eLife.46149.009 We developed an interactive web interface (available at https://vaccinebiomarkers.com) to facilitate data access and further discovery. This website allows users to (1) query genes and visualise their transcriptional profiles for each condition, (2) filter the differentially expressed genes by their functional groups and visualise the fold changes, and (3) analyse WGCNA modules by visualising the functional enrichments and listing the genes of each module. Identification of biomarkers that reflect potential vaccine inflammation In order to reveal potential connections between the different tissues, the differentially expressed genes and each sampling time point, we created a circular heatmap diagram showing genes that were common between the three tissues (Figure 4). We first selected the top 100 genes that were differentially regulated in both the injected muscle tissue and the MLN, then we assessed whether any of these genes were also differentially regulated in the circulating PBMC's RNA expression profiles. This analysis identified a set of genes that coded for soluble or cell-associated proteins present in each of the compartments that were differentially regulated by immunisation. We focused on soluble markers to simplify the sampling and quantification of potential blood biomarkers. The outside rim of the circular heatmap indicates the common individual genes, and the strength of the correlation for each tissue and each immunisation treatment is colour- and sized-coded. (Black indicates positive correlation, red indicates negative correlation; the thickness of the lines corresponds to the magnitude of the correlation coefficient.) Many of these genes encoded chemokines and/or cytokines but the analysis also identified proteins that collectively have been termed acute-phase proteins. These proteins are typically present at high levels during inflammatory events, and they include serum amyloid A-3 (SAA3) and pentraxin 3 (PTX3). Murine SAA3 is an ortholog of the human SAA3 pseudogene and is involved in the murine response to bacterial endotoxins, often acting in combination with TLR2 (Ather and Poynter, 2018; He et al., 2009), whereas the long pentraxin PTX3 facilitates pathogen recognition by macrophages and dendritic cells (Diniz et al., 2004). SAA3 was strongly induced in muscle tissue after most of the immunisations (the exception being Agrippal, whereas in the draining lymph node and the PBMC transcriptome, only LPS, Pentavac SD and poly I:C enhanced RNA expression levels. LPS caused the greatest alteration in gene expression profiles of cyto/chemokine genes in the blood, enhancing CCL2, CCL3, CCL4, CXCL1, CXCL2, CXCL3, CXCL9, CXCL10, CXCL13 and TNF-alpha expression but not affecting CXCL5. These differences were seen mainly at the very early time points of 4 and 8 hr and were not maintained at 24 hr post intramuscular immunisation with the LPS. Notably, in the draining lymph node, the TNF-alpha levels are not elevated. In the muscle, there is strong general upregulation of IL-6 and elevated expression of IL-6 also occurs in the lymph node after LPS and Pentavac SD immunisation, whereas there is a generalised downregulation of IL-6 in the PBMCs. The licensed vaccines, surprisingly including the Pentavac SD which contained the whole cell pertussis, induced minor perturbations of the RNA expression levels in the PBMC. Poly I:C strongly upregulated CCL2 and CXCL10 and mildly enhanced CXCL1, CXCL9 and CXCL13 in the blood, and interestingly this pattern was slightly different from the expression in the draining lymph node where CXCL9, CXCL10 and CCL2 were enhanced while CXCL1 and CXCL13 were unaffected (Figure 4). Figure 4 Download asset Open asset Circular map representing the differentially expressed genes and reporting the observed correlations between each tissue. The three tissues analysed are organised as in Figure 2 and are indicated on the outer part of the circular map. Each outer tile has a different colour to denote which injection they received: Fluad (red), Agrippal (purple), Poly I:C (green), Pentavac SD (yellow), LPS (blue), IFA (grey) and Engerix B (light blue). Each cell in the inner circles shows the fold change for each given time point, treatment, and tissue (dark blue/blue for upregulation and dark red/red for downregulation, with the density of the colour corresponding to the magnitude of fold change). The lines connecting the different subsets of tiles represent the correlation between tissues for each gene-treatment combination with black lines showing a positive correlation and red lines a negative correlation. The thickness of the lines indicates the magnitude of the correlation. https://doi.org/10.7554/eLife.46149.011 Identification of biomarkers in the serum We next examined whether a marker identified from the transcriptomics could be measured in the blood. We measured a panel of cyto/chemokines by Luminex and SAA3 by ELISA in sera harvested from mice vaccinated with saline control, the two licensed vaccines (Pentavac SD and Fluad) and the two potent TLR agonists (LPS and Poly I:C). We selected these treatments on the basis of the range of responses observed from the transcriptomic analysis. Pentavac SD, LPS and Poly I:C had signatures of cytokine responses in the MLN, Fluad much less so, but such responses were still detectable, suggesting that these were good candidates for enabling cyto/chemokine detection in the peripheral blood sera (Figures 2 and 4). Serum was collected from five mice per itreatment and per analysis time point. LPS was clearly inflammatory at early time points after injection, eliciting strong expression of CCL2, CCL3, CCL4, CCL5, CXCL1, CXCL2 and CXCL10.Following treatment with LPS, TNF-α and IL-6 proteins were significantly above baseline and saline control levels at 4 and 8 hr, with a rapid return to much lower expression levels by 24 hr and to basal levels by 48–72 hr (Figure 5A). Levels of CXCL10 were still significantly above those in controls at 24 hr after LPS injection, having reached a peak at 4 hr after treatment which then declined to almost half by 8 hr but was still significantly (p<0.05) above controls and baseline at 503 pg/mL by 24 hr post immunisation. In addition, polyI:C elicited measurably elevated levels of a number of these cyto/chemokines, specifically CXCL10 and CCL5 at 4 hr (CXCL10 – 6,263 pg/mL; CCL5 – 8,727 pg/mL, p<0.0001) and 8 hr (CXCL10 – 1,462 pg/mL; CCL5 – 3,642 pg/mL, p<0.0001) and CCL2 and CCL4 at 4 hr (CCL2 – 3,944 pg/mL; CCL4 – 1,249 pg/mL, p<0.0001). Strikingly, all treatments apart from the saline control elicited very high levels of expression of the SAA3 protein, which were at least 1000-fold greater than those of any other measured analyte, and moreover the kinetics of SAA3 expression were also of a longer duration than those for expression of the cyto/chemokines (Figure 5A). The TLR4 agonist LPS elicited the highest peak SAA3 response of 492.8 µg/mL at 24 hr (p<0.0001), which reduced considerably by 48 hr (99.74 µg/mL, p<0.0001), reaching a level of 5.96 µg/mL by 72 hr and baseline levels after 168 hr. Although the Pentavac SD vaccination did not achieve the same peak level of SAA3 expression as the LPS injection, the levels of SAA3 continued to increase until 48 hr post-immunisation with Pentavac SD (332.6 µg/mL, p<0.0001) and were maintained until 72 hr (245.6 µg/mL, p<0.0001), remaining significantly above baseline levels at the final analysis time point of 168 hr (104.8 µg/mL, p<0.0001). A comparison of the total accumulation of SAA3 after the LPS or Pentavac SD vaccinations revealed that the AUC for Pentavac SD was more than twice that of LPS, being 31,455 µg.hr/mL and 14,572 µg.hr/mL, respectively. The TLR3 agonist Poly I:C invoked an SAA3 expression profile in which the molecule reached 33.11 µg/mL at 4 hr before falling back to 15.04 µg/mL at 8 hr and rising again to 33.54 µg/mL at 24 hr, although these differences did not reach statistical significance when compared to the saline control. Interestingly, treatment with Fluad, which contains the oil-in-water emulsion adjuvant MF59, did not generate a peak in the expression level of SAA3 in the sera until 24 hr post-injection (81.82 µg/mL, p=0.0014), suggesting that a delayed mechanism of action is induced by this emulsion. Figure 5B shows the fold changes over saline alone of measured proteins that are induced by different vaccines at the different time points, giving an indication of the degree and duration of expression over the background levels. In the case of the LPS immunisation, this analysis showed that an ‘expression set’ of cyto/chemokines and SAA3 can be defined to include CCL2, CCL3, CCL4, CCL5, CXCL1, CXCL2, CXCL10, IL-6 and TNF-α but not SAA3 at 4 hr, then the same set of cyto/chemokines but including SAA3 at 8 hr. These comparisons of the differential transcriptomic expression in the blood with the actual levels of expressed proteins measurable in the animal sera revealed that many of the proteins closely matched. The strong expression of CCL2, CCL3, CCL4, CXCL2, CXCL10, and TNF-α proteins that was elicited by LPS immunisation was in line with the measured transcriptomic changes at early time points (4, 8 and 24 hr). By contrast, there is a downregulation in transcript levels for CCL5, but CCL5 protein levels are elevated at 4, 8, and 24 hr. The sustained upregulation of SAA3 following the Pentavac SD vaccination was reflected in both transcript and protein levels. Figure 5 with 1 supplement see all Download asset Open asset Serum chemokine and cytokine expression. Serum chemokines and cytokines were measured using a Luminex bead 9-plex assay, and serum amyloid A3 was measured by quantitative ELISA. (A) Each panel shows the longitudinal expression of an individual chemokine/cytokine in murine peripheral blood sera at times 0, 4, 8, 24, 48, 72 and 168 hr post-injection with the indicated vaccine (n = 5 mice per time point). Chemokine/cytokine levels induced by different vaccines that were in excess of those induced by saline alone were analysed using two-way ANOVA followed by Dunnett’s multiple comparisons test, **p<0.01, ***p<0.001, ****p<0.0001. (B) Volcano plot representing the differential expression of all measured proteins in the sera induced by different vaccines shown as fold-change in the levels induced by saline alone at different time points. Data points corresponding to responses with adjusted p-value <0.01 are labelled with the protein name, and colour coded according to the time point. https://doi.org/10.7554/eLife.46149.012 Correlation between transcripts and circulating cytokines We next quantified the correlations between blood transcript and protein fold changes across all time points for all measured cyto/chemokines in the LPS immunisation group (Figure 5—figure supplement 1). CXC

Systems vaccinology approaches have been used successfully to define early signatures of the vaccine-induced immune response. However, the possibility that transcriptomics can also identify a correlate or surrogate for vaccine inflammation has not been fully explored. We have compared four licensed vaccines with known safety profiles, as well as three agonists of Toll-like receptors (TLRs) with known inflammatory potential, to elucidate the transcriptomic profile of an acceptable response to vaccination versus that of an inflammatory reaction. In mice, we looked at the transcriptomic changes in muscle at the injection site, the lymph node that drained the muscle, and the peripheral blood mononuclear cells (PBMCs)isolated from the circulating blood from 4 hr after injection and over the next week. A detailed examination and comparative analysis of these transcriptomes revealed a set of novel biomarkers that are reflective of inflammation after vaccination. These biomarkers are readily measurable in the peripheral blood, providing useful surrogates of inflammation, and provide a way to select candidates with acceptable safety profiles.

[This corrects the article DOI: 10.1038/s41541-017-0027-3.].

Systems biology has the potential to identify gene signatures associated with vaccine immunogenicity and protective efficacy. The main objective of this study was to identify optimal postvaccination time points for evaluating peripheral blood RNA expression profiles in relation to vaccine immunogenicity and potential efficacy in recipients of the candidate tuberculosis vaccine M72/AS01. In this phase II open-label study (NCT01669096; https://clinicaltrials.gov/), healthy Bacillus Calmette-Guerin-primed, HIV-negative adults were administered two doses (30 days apart) of M72/AS01. Twenty subjects completed the study and 18 subjects received two doses. Blood samples were collected pre-dose 1, pre-dose 2, and 1, 7, 10, 14, 17, and 30 days post-dose 2. RNA expression in whole blood (WB) and peripheral blood mononuclear cells (PBMCs) was quantified using microarray technology. Serum interferon-gamma responses and M72-specific CD4(+) T cell responses to vaccination, and the observed safety profile were similar to previous trials. Two different approaches were utilized to analyze the RNA expression data. First, a kinetic analysis of RNA expression changes using blood transcription modules revealed early (1 day post-dose 2) activation of several pathways related to innate immune activation, both in WB and PBMC. Second, using a previously identified gene signature as a classifier, optimal postvaccination time points were identified. Since M72/AS01 efficacy remains to be established, a PBMC-derived gene signature associated with the protective efficacy of a similarly adjuvanted candidate malaria vaccine was used as a proxy for this purpose. This approach was based on the assumption that the AS01 adjuvant used in both studies could induce shared innate immune pathways. Subjects were classified as gene signature positive (GS(+)) or gene signature negative (GS(-)). Assignments of subjects to GS(+) or GS(-) groups were confirmed by significant differences in RNA expression of the gene signature genes in PBMCs at 14 days post-dose 2 relative to prevaccination and in WB samples at 7, 10, 14, and 17 days post-dose 2 relative to prevaccination. Hence, in comparison with a prevaccination, 7, 10, 14, and 17 days postvaccination appeared to be suitable time points for identifying potentially clinically relevant transcriptome responses to M72/AS01 in WB samples.

The RTS,S candidate malaria vaccine can protect against controlled human malaria infection (CHMI), but how protection is achieved remains unclear. Here, we have analyzed longitudinal peripheral blood transcriptome and immunogenicity data from a clinical efficacy trial in which healthy adults received three RTS,S doses 4 weeks apart followed by CHMI 2 weeks later. Multiway partial least squares discriminant analysis (N-PLS-DA) of transcriptome data identified 110 genes that could be used in predictive models of protection. Among the 110 genes, 42 had known immune-related functions, including 29 that were related to the NF-κB-signaling pathway and 14 to the IFN-γ-signaling pathway. Post-dose 3 serum IFN-γ concentrations were also correlated with protection; and N-PLS-DA of IFN-γ-signaling pathway transcriptome data selected almost all (44/45) of the representative genes for predictive models of protection. Hence, the identification of the NF-κB and IFN-γ pathways provides further insight into how vaccine-mediated protection may be achieved.

In the original version of the article, Fig. 2 b was mislabelled. The second row was incorrectly labelled as “Mi MPL > Mi PBS and Mi QS ≈ M PBS ” this has now been corrected to “Mi QS-21 > Mi PBS and Mi MPL ≈ Mi PBS ”. The third row was incorrectly labelled as “Mi QS21 > Mi PBS and Mi MPL ≈ Mi PBS ” this has now been corrected to “Mi MPL > Mi PBS and Mi QS-21 ≈ Mi PBS ”. These errors have now been corrected in the PDF and HTML versions of this article.

To elucidate the role of innate responses in vaccine immunogenicity, we compared early responses to hepatitis B virus (HBV) surface antigen (HBsAg) combined with different Adjuvant Systems (AS) in healthy HBV-naive adults, and included these parameters in multi-parametric models of adaptive responses. A total of 291 participants aged 18-45 years were randomized 1:1:1:1:1 to receive HBsAg with AS01(B), AS01(E), AS03, AS04, or Alum/Al(OH)(3) at days 0 and 30 (ClinicalTrials.gov: NCT00805389). Blood protein, cellular, and mRNA innate responses were assessed at early time-points and up to 7 days after vaccination, and used with reactogenicity symptoms in linear regression analyses evaluating their correlation with HBs-specific CD4(+) T-cell and antibody responses at day 44. All AS induced transient innate responses, including interleukin (IL)-6 and C-reactive protein (CRP), mostly peaking at 24 h post-vaccination and subsiding to baseline within 1-3 days. After the second but not the first injection, median interferon (IFN)-gamma levels were increased in the AS01B group, and IFN-gamma-inducible protein-10 levels and IFN-inducible genes upregulated in the AS01 and AS03 groups. No distinct marker or signature was specific to one particular AS. Innate profiles were comparable between AS01B, AS01E, and AS03 groups, and between AS04 and Alum groups. AS group rankings within adaptive and innate response levels and reactogenicity prevalence were similar (AS01(B) >= AS01(E) > AS03 > AS04 > Alum), suggesting an association between magnitudes of inflammatory and vaccine responses. Modeling revealed associations between adaptive responses and specific traits of the innate response postdose 2 (activation of the IFN-signaling pathway, CRP and IL-6 responses). In conclusion, the ability of AS01 and AS03 to enhance adaptive responses to co-administered HBsAg is likely linked to their capacity to activate innate immunity, particularly the IFN-signaling pathway.

Significance The RTS,S malaria vaccine is the most advanced malaria vaccine candidate to be tested in humans. Despite its promise, there is little understanding of its mechanism of action. In this work, we describe the use of a systems biological approach to identify “molecular signatures” that are induced rapidly after the standard RTS,S vaccination regimen, consisting of three RTS,S immunizations, or with a different regimen consisting of a primary immunization with recombinant adenovirus 35 (Ad35) expressing the circumsporozoite malaria antigen followed by two immunizations with RTS,S. These results reveal important insights about the innate and adaptive responses to vaccination and identify signatures of protective immunity against malaria.

MotivationExperiments in which the effect of combined manipulations is compared with the effects of their pure constituents have received a great deal of attention. Examples include the study of combination therapies and the comparison of double and single knockout model organisms. Often the effect of the combined manipulation is not a mere addition of the effects of its constituents, with quite different forms of interplay between the constituents being possible. Yet, a well-formalized taxonomy of possible forms of interplay is lacking, let alone a statistical methodology to test for their presence in empirical data.ResultsStarting from a taxonomy of a broad range of forms of interplay between constituents of a combined manipulation, we propose a sound statistical hypothesis testing framework to test for the presence of each particular form of interplay. We illustrate the framework with analyses of public gene expression data on the combined treatment of dendritic cells with curdlan and GM-CSF and show that these lead to valuable insights into the mode of action of the constituent treatments and their combination.Availability and ImplementationR code implementing the statistical testing procedure for microarray gene expression data is available as supplementary material. The data are available from the Gene Expression Omnibus with accession number GSE32986.

Clostridium thermocellum is a thermophilic, cellulolytic anaerobe that is a candidate microorganism for industrial biofuels production. Strains with mutations in genes associated with production of l -lactate ( Δldh ) and/or acetate ( Δpta ) were characterized to gain insight into the intracellular processes that convert cellobiose to ethanol and other fermentation end-products. Cellobiose-grown cultures of the Δldh strain had identical biomass accumulation, fermentation end-products, transcription profile, and intracellular metabolite concentrations compared to its parent strain (DSM1313 Δhpt Δspo0A ). The Δpta -deficient strain grew slower and had 30 % lower final biomass concentration compared to the parent strain, yet produced 75 % more ethanol. A Δldh Δpta double-mutant strain evolved for faster growth had a growth rate and ethanol yield comparable to the parent strain, whereas its biomass accumulation was comparable to Δpta . Free amino acids were secreted by all examined strains, with both Δpta strains secreting higher amounts of alanine, valine, isoleucine, proline, glutamine, and threonine. Valine concentration for Δldh Δpta reached 5 mM by the end of growth, or 2.7 % of the substrate carbon utilized. These secreted amino acid concentrations correlate with increased intracellular pyruvate concentrations, up to sixfold in the Δpta and 16-fold in the Δldh Δpta strain. We hypothesize that the deletions in fermentation end-product pathways result in an intracellular redox imbalance, which the organism attempts to relieve, in part by recycling NADP + through increased production of amino acids.