Study outcomes

The co-primary outcomes of this phase I study were safety, immunogenicity and YF viraemia following vaccination with one of three vYF doses or YF-VAX have been previously published.19 The observational outcomes, reported here, involved the assessment of the inflammatory and innate immune response biomarkers and the cellular immune response following vaccination with one of three vYF doses or YF-VAX.

Luminex for cytokine quantification

All participants provided a pre-vaccination serum sample at baseline (Day 0 [D0]) and post-vaccination serum samples on D1, D3, D5, D7, D10, and D14 for pro-inflammatory cytokine quantification. Serum cytokine concentrations were analysed using two human Milliplex MAP Kits (Merck Millipore, Darmstadt, Germany): the human cytokine/chemokine/growth factor panel A magnetic bead panel for 15 cytokines (IFNγ, TNFα, IL1α, IL1β, IL5, IL6, IL8, IL10, IL12p40, IL17α, IL18, G-CSF, MIP1α, MCP1, IP10 [CXCL10]), and CCL5/RANTES, according to the manufacturer's instructions. The serum samples were added (undiluted for 15-plex and 100-fold diluted for RANTES) to a mixture of colour-coded-magnetic beads, pre-coated with analyte-specific capture antibodies. Streptavidin-phycoerythrin conjugates were added to analyte-specific biotinylated antibodies, and the fluorescent signals were measured using the multiplex array reader Bio-Plex 200 System (Bio-Rad Laboratory, Inc, Hercules, CA, USA). The median fluorescence intensities (MFI) were automatically analysed using a 5-parameter logistic regression standard curve for calculating analyte concentrations in samples (pg/mL). The ratios versus the pre-vaccination time point (D1/D0, D3/D0, D5/D0, D7/D0, D10/D0, D14/D0) were also calculated to evaluate the fold-rise over baseline of the post-vaccination pro-inflammatory response.

Memory B cell IgG and IgM FluoroSpot assay

All participants provided a pre-vaccination blood sample at baseline (D0) and post-vaccination on D28 and on D180 for memory B cell assessments. Memory B cells secreting vYF-specific IgG and IgM antibodies were quantified using a FluoroSpot Assay (FluoroSpot; Mabtech AB, Stockholm, Sweden) according to manufacturer's instructions.

Peripheral blood mononuclear cells (PBMCs) were isolated and cryopreserved until analysis. Cryopreserved PBMCs were thawed and cultured (only samples with viability >70%) for 5 days in RPMI supplemented with Human memory B-cell StimPack (Mabtech AB, Stockholm, Sweden) to induce terminal differentiation into antibody secreting cells (ASCs).

FluoroSpot plates (Mabtech AB, Stockholm, Sweden) were coated with anti-vYF monoclonal antibody HB112-4G2 (in-house mAb; cat# HB-112™; ATCC, Manassas, VA, USA) for vYF-specific responses, or mouse anti-IgG/IgM mAbs (Mabtech AB, Stockholm, Sweden) for total IgG/IgM response; the specificity and reactivity of these antibodies has been previously validated.20^,21 After overnight incubation at 5 °C, washing and blocking, the vYF vaccine was then added to the vYF-specific wells (10.76 Log GEq/mL, Genomic Equivalents per mL), except for negative controls (HB112-4G2 mAb only) and incubated for 4 days at 5 °C. Then, serial dilutions of stimulated PBMCs were added to vYF-specific wells, or to total IgG wells, for 5 h and removed. The following day, plates were incubated for 2 h at 20 °C with respective fluorochrome-labelled detection antibody anti-IgG 550 (clone MT78/145; Human IgG FluoroSpot FLEX kit, cat# FSX-05R-10, Mabtech AB, Stockholm, Sweden) or anti-IgM 640 (clone MT22; Human IgM FluoroSpot FLEX kit, cat# FSX-17M, Mabtech AB, Stockholm, Sweden). The plates were finally washed and dried at ambient temperature (protected from light), before counting IgG and IgM spots (ASCs) with the FluoroSpot reader (ISpotSpectrum, AID, Strassberg, Germany).

For each sample and each isotype, vYF-specific ASC/10⁶ cells (after subtraction of 4G2 mAb ASC/10⁶ cells) and total ASC/10⁶ cells were calculated to assess the frequency of IgG and IgM vYF-specific responses. The vYF-specific responses were calculated by dividing the respective number of vYF-specific IgG or IgM ASC/10⁶ cells by the total IgG or IgM ASC/10⁶ cells. The D28/D0 and D180/D0 ratios were calculated to evaluate the fold-rise over baseline of the specific and total responses at 28- and 180-days post-vaccination, respectively.

Whole blood RNA transcriptomic analysis of selected genes by RT-qPCR

The expression of 85 genes (Supplementary Table S1) previously shown to be involved in the immune response to YF-17D vaccination in NHPs18 (and humans22, 23, 24, 25), as well as 5 housekeeping genes (NONO, GUSB, SDHA, HPRT1, and UBE2D2), were assessed by RT-qPCR assay on the Fluidigm digital array platform using a customised set of 90 TaqMan Gene Expression Assays (Thermo Fisher Scientific, Waltham, MA, USA) for the corresponding selected genes (Supplementary Table S1). The criteria used to select this panel of genes for immuno-profiling by RT-qPCR were: 1) genes clearly involved in the immune response based on the literature (all genes with unknown or unclear function were not selected); 2) NHP genes without known orthologues in humans were not selected; 3) genes up or down regulated in both humans and NHPs or in human only (to avoid elimination of human differentially expressed gene [DEG] since this was our final target, earlier kinetics are observed in NHPs; we may have missed the timepoint where adaptative response initiation was observed [D14 in human]); 4) all the genes differentially expressed only at D21 and/or D28 post-vaccination were not selected.25 For the later criteria, at D21 and/or D28 post-vaccination timepoints, the activated pathways were not considered linked to vaccination (these likely reflect renewal of erythrocytes following frequent blood sampling). In addition, the genes that encoded the heavy constant chain of immunoglobulins IgG1, IgG2, IgG3, IgG4, and IgM were added to the list of genes selected to assess their expression level as these were previously shown to be involved in the immune response to YF vaccination and could be a marker of B cells and humoral response.24 Although numerous potential immunoglobulin genes were identified in the BioVacSafe human study based on the annotations of the probes identified as DEGs at D14,25 the selection was focused on heavy constant chains to maximise the chance of detecting the expression of immunoglobulin while reducing the number of targets to design. TLR3 was also identified in all NHPs but not in humans. However, since it was identified as a key component of the immune response to the vaccine in NHPs, we decided to add it to the panel.

In brief, whole blood was collected in PAXgene tubes (PreAnalytix, Qiagen-BD Company, Hombrechtikon, Switzerland) at baseline and at D1, D3, D5, D7, D10, D14, D28, and D180, and stored according to the manufacturer's protocol prior to RNA extraction. Total RNA was extracted with the PAXgene 96 Blood RNA kit (PreAnalytix, Qiagen-BD Company, Hombrechtikon, Switzerland) according to manufacturer's instructions on the Tecan Evo 150 workstation (Tecan, Männedorf, Switzerland). RNA samples were quantified by spectrophotometry and stored at −80 °C until further processing and assessment. Potential residual genomic DNA in RNA samples was eliminated by a Turbo DNase step followed by an RNA clean-up and concentration step on silica column (Clean and concentrator; Zymo Research, Irvine, CA, USA). The RNA quantity, purity, and quality were checked on the TapeStation (Agilent Technologies, Santa Clara, CA, USA) and the DropSense 96 or Lunatic (Unchained Labs, Pleasanton, CA, USA). Samples with A260/A280 ratio ≈2, RNA Integrity Number (RIN) ≥6, and minimal RNA concentration of 50 ng/μL were used for the next steps. Residual genomic DNA contamination was quantified by qPCR Femto™ assay (Zymo Research, Irvine, CA, USA) and had to be <5pg/μL after the final dilution.

The RNA extracted from the samples was reverse-transcribed using the Reverse Transcription Master Mix kit (Standard Biotools, San Francisco, CA, USA) according to the manufacturer's instructions. Each cDNA sample was preamplified using TaqMan PreAmp Master Mix 2X kit (Applied Biosystems, Bedford, MA, USA) and a pool of the 90 TaqMan pairs of primers to increase the number of copies of target DNA. The qPCR step was performed in a Fluidigm BioMark qPCR thermocycler (Standard Biotools, San Francisco, CA, USA) with a TaqMan Universal PCR Master Mix kit (Applied Biosystems, Bedford, MA, USA). For each sample, two independent RT-qPCR runs were performed. At each step, we included positive and negative controls. Positive controls were prepared using blood from a flavivirus-naïve subject (Blood donation from Etablissement Français du Sang [EFS], France), and negative control consisted of nuclease-free water. We also included: extraction positive (TextP) and negative (TextN) controls, RT step positive (TrtP) and negative (TrtN) controls, preamplification positive (TpAmpP) and negative (TpAmpN) control, PCR non-template control (NTC), PCR non-reagent control (NRC, without the detection probe), and no RT control (TnoRT).

C_t values were computed using Fluidigm Real-Time PCR analysis software (Standard Biotools, San Francisco, CA, USA) and normalised using the fluorescence of ROX passive reporter (Raw Taqman—Background Taqman)/(Raw ROX—Background ROX). Fluidigm Biomark™ quality thresholds, based on amplification curves, were used at the default 0.65 threshold value. Overall, only 0.2% of C_t values located outside the C_t optimal 6–25 range recommended by Fluidigm were considered undetectable and coded at 25 (no values observed <6). Almost all NRCs and other negative control samples (NTC, TpAmpN, TrtN, TextN, TnoRT) were undetectable, and almost all positive controls (TpAmpP, TrtP, TextP) were detected, except for 11 target genes (CCL2, IFI6, IGHG3, IRF7, ISG15, KCTD14, LY6E, PARP9, SDHA, TRIM22, XAF1) having more than 25% flagged reactions in at least one plate that were excluded from all analyses. Missing values for 33 flagged reactions were imputed using the target mean.

Thirty-five out of 595 samples were not analysed because they were either lost (broken tubes [n = 6]) or had poor quality mRNA (RNA concentrations <50 ng/μL, TapeStation RINe <6 [n = 29]). For samples having a genomic DNA (gDNA) quantity over the threshold based on Femto qPCR assay (i.e. >5 pg/μL after the final dilution step), an additional control was added by testing the RNA directly without the reverse transcriptase (RT) step (no RT control) to ensure that the residual gDNA was not detected with the different qPCR reactions. Samples from one participant with a very high proinflammatory profile at baseline were excluded as outliers. This participant also showed high IP10 [CXCL10] concentration in serum at baseline. No sample (other than that mentioned) appeared as a potential outlier, before and after normalisation, when checked by scaled principal component analysis (PCA) and hierarchical clustering using all the target genes. The quality control was completed by density and correlation plots of all samples.

A two-step delta–delta C_t (DDCt) normalisation was applied to each assay plate, using 4 of the 5 predefined housekeeping genes (these genes showed only limited variation at only one time point) and 3 human sample TrtP controls (these control samples were not different from those of study participants at baseline and clustered with them). The assay plates were replicated in pairs with high reliability (intraclass correlation coefficient >0.9 in a prior technical pilot study), batch corrected using the Combat parametric algorithm (sva bioconductor package, version 3.42.0)26 and summarised by the mean of the pair of values, except for one sample (participant #19 baseline sample) where the first run (but not the second) gave very different results from those of the other participants, suggesting a potential issue. As the gene expression for each time point for each participant was normalised compared to their baseline, to avoid introduction of a “run effect” bias, all time points for the same participant were assessed on the same plate (i.e. treated at the same time, with the same operator and same reagent batches).

Ethics

The study was conducted in compliance with the International Conference on Harmonisation guidelines for Good Clinical Practice and the principles of the Declaration of Helsinki, as well as local and/or national regulations and directives. The protocol and amendments (WRAIR 2689, HRPO Log E00378.6a-EID001) were approved by the WRAIR Institutional Review Board and applicable local and/or national regulatory agencies. Written informed consent was obtained from participants before study enrolment.

Statistical analysis

All analyses for this publication were performed on the full analysis set (FAS). The FAS consisted of all participants who received one dose of vYF or YF-VAX and had at least one valid post-vaccination blood sample result.

vYF-specific ASC frequencies were evaluated by descriptive statistics at the timepoints assessed: geometric means (GM) of vYF-specific IgG and IgM ASC frequencies, respective geometric mean ratios (GMR) at each post-vaccination timepoint versus pre-vaccination (D0) and associated 95% CIs.

Time course inferential statistical analyses for differences between vaccines and vaccine doses were implemented using the empirical Bayesian approach of Ritchie et al. 2015 (limma Bioconductor package, version 3.50.3),27 which provides more robust estimates of DEGs by borrowing the gene variance estimation across all the gene targets. Repeated time point measurements for participants were handled according to the usual method of the limma package.

Normality of the DDCt values was tested by group and gene using Shapiro Wilk test leading to a majority of non-significant p values (additionally the whole nonparametric analyses were performed showing similar test conclusions compared to their parametric counterparts). Two series of longitudinal analyses were presented to handle possible effects of baseline imbalance between vaccines and dose groups: absolute time points and change versus baseline (D0). Group differences, for both absolute change and change against baseline series, were tested by adding the interaction term “group with time point” into the model, and by further testing adequate contrasts: e.g. (DDCt D7—DDCt D0)_YF-VAX—(DDCt D7–DDCt D0)_{vYF_all_doses} for the change from baseline (second series). To correct for multiple testing in the linear model comparisons, Benjamini and Hochberg's False Discovery Rate (FDR) procedure was applied. The inferential analysis results were presented using heatmaps with hierarchical clustering, and additionally with individual participant time course curves.

PCA was computed using prcomp R function, and scaled as gene expression exhibits various ranges, and hierarchical clustering was computed using complete linkage method with Euclidean distance. UpSet plot was produced using R package ggupset version 0.2.1.

Systemic cytokines/chemokines

After vaccination with vYF (4, 5, and 6 Log CCID₅₀) or YF-VAX there was no increase in the serum concentration of the 16 cytokines/chemokines assessed through to D14 (Supplementary Table S2), except for IP-10 (CXCL10), where an approximate 2-fold increase was observed from D1 to D7 followed by a decrease to baseline by D14 (Fig. 1). The changes in IP-10 levels were not significantly different between the vYF groups or compared with the YF-VAX group.

Open in a separate window

Fig. 1

Serum IP-10 (CXCL10) concentration following vaccination with vYF (4, 5 and 6 Log CCID₅₀) and YF-VAX (FAS; n = 72). Data shown for individual participants, with the geometric mean indicated by the black line, and lower and upper black lines representing the 95% confidence interval.

Targeted transcriptomic responses

The expression profiles of the genes involved in early immune responses to vaccination with vYF (irrespective of dose) and YF-VAX were similar (Fig. 2). The innate immune response signatures (antiviral responses [type 1 IFN, IFN signal transduction; IFN-stimulated genes], activated dendritic cells, and viral sensing pattern recognition receptors) peaked at D7 across the vYF and YF-VAX groups. Those for the adaptive immune response (cell division in stimulated CD4+ T cells [CDT1, BIRC5, and KIAA0101 (PCLAF)], B cell and antibody response [IGHM, IGHG1, MZB1, and TNFSF17]) peaked at D14. One participant from the YF-VAX group was excluded from the statistical analysis on a biological basis, as they exhibited a very high inflammatory response at baseline. Nevertheless, this participant's immune response after vaccination followed a similar expression profile as the others, with activation of the same pathways.

Open in a separate window

Fig. 2

UpSet plots of differentially expressed genes for each time point versus the baseline, and the number of genes in common for each comparison across the vaccine groups (FAS; n = 72) (a). Genes are considered in common if they have the same fold change direction. In the bar plot, red represents up-regulated and green down-regulated genes. Heatmaps of selected differentially expressed genes after vaccination at each time point versus baseline, grouped by functional category: (b) antiviral innate response (type I interferon signature, IFN signal transduction/ISG); (c) adaptive immune response (cell division of stimulated CD4+ T cells, B cells and antibody response); (d) viral sensing; and (e) activated dendritic cells.

The expression kinetics of the selected DEGs relative to baseline are presented as heatmaps in Fig. 2b for between vaccine differences and between vYF dose differences, respectively. The significant gene expression changes from baseline for all groups appeared similar (same up- or down-regulated DEG profiles), with no significant (absolute and versus baseline) differences between groups in nearly all comparisons. Supplementary Figure S1 shows the individual participant time course for IFI27 gene expression by vaccine and dose group for illustration purpose. The time course of all genes expressed was assessed, and the IFI27 gene is presented as an example, as it presented the highest fold change. The UpSet plots in Fig. 2a summarise the overlap of gene expression changes shared by the vaccine groups across time points. It shows that a majority of DEGs were up-regulated and their response was maintained for all timepoints between D3 and D10. Although no significant differences appeared between groups, a slight kinetic difference was observed in the peak responses (D5 or D7) (Fig. 2b and Supplementary Figure S1).

The authors thank Julie Piolat and Yichen Jia for statistical analysis. The authors also thank the staff of the WRAIR Clinical Trials Center, Emerging Infectious Diseases Branch and the Henry Jackson Foundation (Cooperative Agreement #W81XWH-07-2-0067 and Diagnostics and Countermeasures Branch [DCB] Support Services contract #W81XWH-18-C-0337) for supporting the execution of the vYF study (WRAIR protocol # 2689 EID 001) and particularly the study participants who volunteered for this first-in-human phase I clinical trial.

Editorial assistance with the preparation of the manuscript was provided by Richard Glover, inScience Communications, Springer Healthcare Ltd, London, and was funded by Sanofi. The authors also thank Anirban Sanyal, PhD, for editorial assistance and manuscript coordination on behalf of Sanofi.