To investigate host responses to live-attenuated oral typhoid vaccines, we vaccinated healthy adult volunteers with either three doses of the licensed vaccine Ty21a (n = 33), a single dose of the candidate vaccine M01ZH09 (n = 33), or placebo (n = 33), as part of a randomized, double-blinded, placebo-controlled trial (24). Gene expression (GEX) profiles were generated based on mRNA purified from whole blood before and at multiple time points following vaccination. The majority of measurements were using samples taken 7 days after completion of vaccines, i.e., at D-21, 7 days after the third dose of Ty21a and the first dose of M01ZH09 or placebo (Figures 1A,B).

At D-21, we observed moderate differential expression of the blood transcriptome in Ty21a vaccine recipients (90 genes) relative to baseline, and an even more modest differential perturbation in M01ZH09 vaccine recipients (70 genes, p < 0.001; Figures 1C,D). Due to the low level of perturbation in placebo recipients (29 genes; Figure S1A in Supplementary Material), subsequent analyses focused on Ty21a and M01ZH09 groups only. Transcriptional perturbation at early time points following vaccination (Ty21a: “D-28” and “D-26”; M01ZH09: “D-24”) was more pronounced in Ty21a recipients, potentially driven by the sequential doses of Ty21a in comparison with the single dose of M01ZH09. While we observed only one gene that overlapped both time points following Ty21a vaccination, more overlap was observed between D-24 and D-21 following M01ZH09 (Figure 1E). Interestingly, although Ty21a seemed to induce a similarly modest magnitude of transcriptional changes, it was significantly less immunogenic, as determined by anti-O9:LPS and anti-H responses 28 days following vaccination than was M01ZH09 (Figure S1B in Supplementary Material) (24).

To interrogate the signaling pathways stimulated by vaccination, we performed pathway overrepresentation analysis (ORA) of genes differentially expressed (DE) compared with baseline using the publically available database InnateDB (25). For Ty21a recipients, this analysis highlighted differential regulation of the pathways involved in the adaptive immune system and mRNA splicing. Furthermore, more general signatures of ribosomal subunit activity and translation elongation were highly overrepresented at the early D-26 time point and decreased at the D-21 time point. While ribosomal, and translation and elongation pathways were also DE soon after M01ZH09, we observed strong differential regulation of several pathways associated with cell cycle activity at 7 days following vaccination (D-21, Figure 1F). Although pathway analysis suggests differential activation of distinct molecular patterns compared between the vaccine arms, other similarities in expression patterns were observed, specifically at the earlier time points (D-28, D-26, and D-24) following vaccination.

To further interpret the overall molecular pattern underlying the GEX following live-attenuated oral vaccination against typhoid, we used Gene Set Enrichment Analysis (GSEA). GSEA is a complementary method to pathway ORA that instead of relying on threshold values to identify over- or underexpressed pathways or modules, employs a ranked list of genes to calculate enrichment scores for modules of interest (26). GSEA was performed using 14,302 genes ranked by log₂ fold change(FC) against a conceptual framework of predefined blood transcriptional modules (BTMs) previously described by Li et al., with calculation of a normalized enrichment score (NES) and significance threshold for each module (27).

Shortly after vaccination with two (D-28) or three (D-26) doses of Ty21a, significant perturbation of several BTMs was observed (BH-adjusted p < 0.05) (Figure 2A). While some BTMs representing monocyte (M11.0 and M118.0) and innate response signatures (M.37.0, M16, and M25) were downregulated, several interferon (M127, M111.1, and M75), NK cell (S1, M61.0, M61.2, and M7.2), and T cell (M7.0, M7.3, and M223) modules were upregulated immediately following Ty21a ingestion (Figure 2A). Vaccination with M01ZH09 induced a higher number of BTMs at both time points measured (D-24 and D-21), compared with any time point after Ty21a vaccination. In addition, despite similarities in expression of some cell cycle and IFN modules, M01ZH09 elicited several responses at D-24 that contradicted those seen at the equivalent earliest time points after Ty21a receipt (D-26). These represented upregulation of monocyte (M11.0 and M118.0) and innate response (M37 and M16) modules and downregulation of T cell modules (M7.0, M7.3, and M223) in the M01ZH09 arm (Figure 2A).

Comparison of BTMs 7 days after completion of vaccination (D-21 in both groups) indicated further distinct differences between responses to the two oral vaccines. While positive enrichment of BTMs reflecting cell cycle control and monocytes was measured following M01ZH09, the opposite was observed (with negative or lack of enrichment) after Ty21a vaccination (BH-adjusted p < 0.05, Figure 2B—middle and left). By contrast, modules reflecting NK cell activity were upregulated after Ty21a and downregulated after M01ZH09 vaccination (BH-adjusted p < 0.05) (Figure 2B—middle and right). Of note, while most T cell modules were downregulated at D-21 following M01ZH09 receipt (and not significantly enriched following Ty21a), possible important differences were seen in three T cell-related modules. These modules [“mitotic cell cycle in stimulated CD4 T cells (M4.5),” “mitotic cell cycle in stimulated CD4 T cells (M4.9),” “cell division (stimulated CD4+ T cells) (M46)”] were upregulated in the M01ZH09 group but downregulated in the Ty21a group and represent cell division in CD4⁺ T cells indicating possible activation of specific T helper responses specifically following M01ZH09 vaccination. Overall, this group level GSEA has demonstrated interesting similarities and differences in the transcriptional response to two oral, live-attenuated typhoid vaccines. While transcriptional signatures of T cell responses appeared to be variable, a consistent difference in the expression of signatures representing NK cells was observed between the Ty21a and M01ZH09 arms.

To further analyze the expression dynamics of modules among participants, we performed single sample GSEA (ssGSEA) at day 7 after vaccination (D-21). Plotting the mean expression (mean + SEM) of BTMs reflecting the cell cycle, NK cells, T cells and B cells highlights the differential expression pattern of cell cycle and NK cell modules between the vaccine groups (Figure 3A). Host immune responses to stimuli such as vaccines are known to vary between individuals as a result of both environmental and genetic factors (28, 29). To assess this variability, we plotted NK modules for each participant following vaccination and calculated the inter-decile range for each participant. This analysis highlighted the considerable variability in the response among the participants (Figure 3B). Overall, however, we observed consistent opposite expression of genes in the NK cell BTM including inhibitory and activating receptors including KIR3DL3, KIR3DL1, KIR2DL3, and KIR2DL4 (Figures 3C,D).

Because ssGSEA provides enrichment scores for each BTM in each participant, these can be used to relate BTM expression to immunogenicity measurements (i.e., antibodies 4 weeks following vaccination) as well as parameters following challenge. Using Spearman’s rank correlation, we observed significant associations of cell cycle BTMs expressed at D-21 with antibody responses 28 days after vaccination, which were positive in M01ZH09 but negative in Ty21a recipients (Figure 3E). Moreover, we observed positive correlation of BTMs representing inflammation, monocytes and DCs and anti-H responses following Ty21a compared with M01ZH09 (Figure 3E). In addition, T cell-related modules were negatively associated with antibody responses following Ty21a vaccination. Of note, modules associated with B cell signaling and antigen presentation were negatively associated with serological responses to this vaccine. These data provide further insight into the modular response 7 days after vaccination, and the relationship to humoral responses suggesting that cell cycle BTMs may be predictive of humoral immunogenicity following oral live-attenuated vaccination.

Since this study was performed as part of a human challenge model, we selected two outcome measures following challenge to be correlated with enrichment scores at D-21 following vaccination. Time to diagnosis (ttDx) was delayed in participants receiving the M01ZH09 vaccine whereas there was no difference in maximum temperature (Temp) within 14 days following challenge in participants of both vaccine arms (24). This analysis showed that several modules were associated with post-challenge parameters, including “CD28 co-stimulation” (M12) and “enriched in cell cycle” (M167) following M01ZH09 vaccination. By contrast, following Ty21a vaccination, we observed BTMs that relate to transmembrane transport (M154.1), amino acid transport (M154.0), SMAD2/3 signaling (M97), E2F transcription factor (M8), and complement activation (M112.1) associated with post-challenge parameters. Associations with Temp and ttDx were inverse for several modules within respective vaccine arms (Figure 3E—gray nodes), showing a positive correlation with Temp but negative correlation with ttDx at the same time. Thus, M112.1, M97, and M8 expression following vaccination were positively associated with higher Temp, but shorter ttDx. By contrast, M12, M154.1, and M154.0 expression following vaccination was associated with lower Temp and longer corresponding ttDx (Figure 3F). These data underline the inverse relationship between disease severity markers (Temp) and ttDx of overt clinical disease in the challenge model and indicate possible protection indicated by module M154.0, M154.1, and M12 expression at time point D-21 following vaccination.

Given the major differential response in NK cell module activation between Ty21a and M01ZH09 vaccine recipients, we hypothesized that these vaccine strains might display differences in their capacity to activate NK cells in vitro. In vitro infection of PBMCs isolated from vaccine-naïve study volunteers (n = 16) before vaccination with either of the two vaccine strains (Figure 4A) indicated that Ty21a displayed greater capacity to induce surface expression of NK cell functional activation markers, including CD25 (p < 0.01) and CD107a (p < 0.0001), but not intracellular production of IFN-γ (p = 0.08), compared with M01ZH09. We also observed activation of these markers by Ty21a in CD3⁺ T and CD3⁺CD56⁺ (NKT and γδT) cell populations; however, no differences were observed between the two vaccine strains for CD107a and CD25 surface expression (Figure 4B).

Following vaccination with specific antigens, NK cells have been shown to mount a rapid effector recall response when re-exposed to the same antigen (30). To evaluate this phenomenon in the context of live-attenuated oral vaccines against typhoid, we reexposed PBMCs collected following vaccination (D-14 and D0) to the autologous vaccine strain in study participants. Overall, recall responses on exposure to the autologous vaccine strain did not differ at 14 days (D-14) or 28 days (D0) after vaccination with Ty21a or M01ZH09 (Figure 5). While some participants in the M01ZH09 arm appeared to display an enhanced response to the in vitro stimulation of post-vaccine PBMCs, the recall responses were not significantly increased compared with stimulated PBMCs collected from Ty2a vaccine recipients. Similar observations were made when PBMCs from participants were reexposed to the heterologous vaccine strain: no significantly increased recall responses were observed when PBMCs isolated from M01ZH09 and Ty21a recipients to Ty21a or M01ZH09, respectively (Figure S2A in Supplementary Material). Finally, we exposed the PBMCs to S. Typhi Quailes strain, the strain used in the subsequent challenge (24). Similar to the previous re-exposure experiments, a recall response to this strain was not observed after vaccination (Figure S2B in Supplementary Material).

A randomized, double-blind, placebo-controlled trial was performed at the Centre for Clinical Vaccinology and Tropical Medicine (Churchill Hospital, Oxford, UK) to assess the protective efficacy of a single dose of M01ZH09 (oral suspension and liquid formulation) compared with placebo against S. Typhi challenge 28 days after vaccination (NCT01405521), as previously described (24). Three doses of Ty21a (enteric-coated capsules) were given to a positive comparator arm. The trial was sponsored and monitored by the Oxford University Clinical Trials and Research Governance office, approved by NRES South Central—Oxford A (11/SC/0302) and conducted in accordance with the principles of the International Conference of Harmonization, Good Clinical Practice guidelines.

After study initiation (November 2011), an independent Data and Safety Monitoring Committee (DSMC) reviewed clinical and laboratory data relating to patient safety (months 1, 5, and 8) and interim unblinded analyses of vaccine efficacy (months 5 and 8). No changes to the study protocol or participant eligibility were recommended (24). Four weeks following vaccination, participants ingested 1.2 g/120 mL NaHCO₃[aq] followed by 30 mL of 1–5 × 10⁴ CFU of S. Typhi Quailes strain using an ambulant study design (23). Following challenge participants were followed over 14 days, after which a 2-week course of antibiotics (ciprofloxacin, 500 mg twice daily, 14 days) commenced unless participants developed overt clinical disease, in which case antibiotic treatment commenced immediately. Criteria for typhoid diagnosis were either microbiological (≥1 positive blood culture collected after day 5) and/or clinical (fever ≥38°C sustained for ≥12 h) (23, 24).

Peripheral venous blood (3 mL) was collected in Tempus blood RNA tubes (Applied Biosystems) at baseline (pre-vaccination controls) (n = 66) for all participants. For individuals randomized to receive Ty21a (three-dose schedule) blood was collected at 4 (D-28; n = 9) and 6 days (D-26; n = 9) after the first dose of Ty21a, and 7 days following the final (third) dose of the Ty21a vaccine (D-21; n = 20). For individuals randomized to receive M01ZH09 (single dose), blood was collected 4 days (D-24; n = 10) and 7 days (D-21; n = 30) following M01ZH09 vaccination (Figure 1A). Samples from the placebo arm were also collected at baseline (D-28; n = 8), 4 days (D-24; n = 9), and 7 days (D-21; n = 8) thereafter. Total RNA was extracted from all samples using the Tempus™ Spin RNA Isolation kit (Life Technologies). 50 ng of RNA was used for hybridization into Illumina HT-12v4 bead-arrays (Illumina Inc.) at the Wellcome Trust Sanger Institute, UK, and fluorescent probe intensities captured with GenomeStudio software (Illumina Inc.).

Differential gene transcription was assessed by analysis of changes in transcript abundance in whole-blood samples collected during the study. Raw data were quantile normalized, background subtracted and log₂-transformed using the Bioconductor suite in R (Version 3.0.1) (49). Transcripts represented by more than one probe on the array were collapsed using the probe with the highest mean expression for each gene using the collapseRows() function of the WGCNA package in R (50). The collapsed dataset was then filtered to include only probes significantly detected (p < 0.05) in 60% of all samples, yielding 14,302 genes. This preprocessed dataset was used for all subsequent analyses. Quality control was performed using the arrayQualityMetrics() function in R (51) and led to the exclusion of 14 samples. DE genes were determined by fitting a linear regression model including post-vaccination time point and date on which the array was run as covariates, and using participant number as blocking variable to ensure pairwise analysis using the limma-package in R (52). Contrasts were generated between each time point and pre-vaccination baselines.

Expression patterns following vaccination were analysed using GSEA in R. Gene lists, consisting of 14,302 genes derived as described in the Section “Data Processing and Identification of Differentially Expressed Genes,” were ranked by average FC and applied to the GSEA algorithm (53). Briefly, NESs of BTMs previously described by Li et al. (27) were calculated by evaluating the overrepresentation of module genes at the extremes (i.e., top or bottom) of the ranked gene list and nominal p-values calculated using a permutation test procedure (26). As this method can be applied to any ranked gene list, we adapted the analysis to calculate enrichment scores on a sample-by-sample basis (ssGSEA). In this analysis, per participant FCs at D-21 were calculated using each participant’s respective pre-vaccination baseline sample. Resulting lists was then ranked by FC and applied to the GSEA algorithm. BTMs, which were significantly enriched (p < 0.01) in ≥40% of the samples and only annotated BTMs were included in subsequent analyses.

Vaccine responses were determined by measuring anti-LPS (O9:LPS) and anti-flagellin (H-antigen) IgM, IgA, and IgG antibodies using ELISA before and 28 days after vaccination, as previously described (24).

Peripheral blood mononuclear cells were isolated from study participants pre-vaccination (D-32 Ty21a; D-28 M01ZH09), and at 14 (D-14) and 28 (D0) days after vaccination and stored in liquid nitrogen. For further analysis, samples were thawed, seeded at 2 × 10⁵ cells/well in 96-well tissue culture plates in RPMI + 10% fetal bovine serum (FBS) and rested for 2–3 h. Frozen sucrose stocks of Ty21a and M01ZH09 grown to log-phase in Luria-Bertani broth were thawed and bacteria washed twice in phosphate-buffered saline before resuspension in RPMI + 10% FBS. Bacterial suspensions were added to PBMC in vitro at a multiplicity of infection (MOI) of 0.1:1. A low MOI was chosen to avoid maximal responses in order to interrogate increased recall responses following stimulation of PBMCs from vaccinated individuals. Gentamicin was added at 200 µg/mL 1 h post-infection to kill extracellular bacteria. Recombinant human (rh) IL12 (5 ng/mL) and rhIL18 (50 ng/mL) were used as a positive control. Following addition of bacteria or cytokines, PBMCs were cultured at 37°C, 5% CO₂ for 15 h before addition of Golgistop and Golgiplug according to the manufacturer’s directions, before further culturing for 3 h.

Peripheral blood mononuclear cells were washed and stained with antibodies against cell surface markers CD3 (BD Biosciences), CD56 (BioLegend), CD107a (BD Biosciences), and CD25 (eBioscience). Subsequently cells were fixed and permeabilized (BD Cytofix/Cytoperm) before intracellular staining of IFNγ (BioLegend). Flow cytometric analysis was performed on a BD FACSAria flow cytometer collecting an average of 60,354 events (minimum: 50,000). Gating of cell populations was performed as follows: single events → live cells → lymphocytes → T cells (CD3⁺CD56⁻), NKT cells (CD3⁺CD56⁺), and NK cells (CD3⁻CD56⁺). Expression of surface markers CD107a and CD25 was determined for each cell population through comparison with FMOs run in each experiment. Expression of the intracellular cytokine IFNγ was determined for each cell population through comparison with a sample stained with an isotype control antibody run in each experiment (Figure S3A in Supplementary Material). For each cell subpopulation, at least 102 events were collected (Figure S3B in Supplementary Material). All data analysis was performed using FlowJo v10 (FlowJo, LLC).

To identify modules associated with immunogenicity and parameters following challenge, we performed Spearman’s rank correlation relating NESs for each participant and module to antibody responses 4 weeks following vaccination and maximum Temp within 14 days after challenge. For ttDx was defined as hours from challenge until antibiotic treatment commenced and participants not reaching the diagnostic criteria following challenge were censored at 337.2 h (14.05 days) after challenged. All statistical analysis was performed in R (version 3.2.4) (54) and association networks visualized in Cytoscape v3.2.5 (55).