Significant changes in gene expression precede the robust induction of anti-stalk antibodies
This clinical trial was designed under the assumption that adult humans possess pre-existing immunity to the H1 HA including low levels of antibodies and memory B cells with specificity against the HA stalk domain. The objective was to redirect the immune response to the immunosubdominant stalk through sequential vaccination with cHA constructs that feature head domains from avian influenza virus subtypes but share the same H1 stalk domain (Figure 1A). Fifty-three volunteers were randomized into three different vaccine groups and two placebo control groups in a regimen of prime-booster vaccination: Group 1 (G1) received cH8/1N1 live attenuated influenza vaccine (LAIV) on day 1 followed by AS03A-adjuvanted cH5/1N1 inactivated influenza vaccine (IIV) on day 85 (LAIV8-IIV5/AS03). Group 2 (G2) received the same vaccination regimen but with nonadjuvanted booster vaccination (LAIV8-IIV5). Group 4 (G4) received adjuvanted cH8/1N1 IIV followed by adjuvanted cH5/1N1 IIV (IIV8/ AS03-IIV5/AS03). Group 5 (G5) served as outpatient placebo group and received PBS intramuscularly twice (PBS-PBS). The sequential vaccination strategy as well as the scheme of vaccination groups and blood collection timeline is shown in Figure 1A-C. A detailed description of the trial design, immunogenicity and safety of the experimental vaccines can be found in Nachbagauer et al6. Of note, an additional saline intranasally placebo group was enrolled (G3) but was not included for the present RNAseq analysis because it was considered equivalent to G5.
Serum samples for antibody quantification were collected at baseline (day 1), and day 29 post-prime; day 85 (pre- boost) and day 113 (day 29 post- boost) (Figure 1C). We used data from Nachbagauer et al6 to represent the anti-stalk antibody induction shown in Figure 1D-E. The group receiving the IIV8/AS03 priming (G4) had the higher increase in stalk antibodies while priming with LAIV8 did not induce specific antibodies against the stalk of the HA. Comparable results between the groups were found after the boost, with higher induction in subjects from G1 receiving IIV5/AS03 as a booster, followed by volunteers from Group 2 (IIV5 booster). To investigate early molecular changes after receiving the experimental vaccines, whole blood was subjected to RNA sequencing for transcriptional profiling and the correlation of these changes with the anti-stalk antibody responses found in Nachbagauer et al6 investigated. For this, blood was collected at baseline (day 1), day 3, day 7 and day 29 post prime; and day 85 (pre- boost), day 93 (day 7 post- boost) and day 113 (day 29 post- boost).
We first performed differential gene expression analysis to assess the dynamic changes in gene expression between each of the vaccination groups compared to placebo. A total of 227 genes were differentially expressed between the vaccination groups and placebo for all the timepoints, with no significant differences at baseline. A heatmap of average gene expression changes for each comparison indicated that most changes occurred at day 3 and day 7 in the G4 group (IIV8/AS03 – IIV5/AS03), while significant differences were absent for the other vaccination groups after the prime (Figure 2A). These responses decreased by day 28 and pre-boost, at day 85. Additionally, we found significant differentially expressed genes (DEGs) for G1 seven days after receiving the booster dose on day 92. The top G4 induced genes at day 3 included genes related to innate immunity activation and IFN signaling while DGEs at day 7 were related to B cell proliferation signatures. Interestingly, the transcriptional profile was similar between these two groups after receiving the IIV/AS03 vaccine for the first time, marked by expression of IGHG1, IGLV1-44 and B cell related genes at day 7 post- prime for G4, and day 7 post-boost in G1.
Next, we performed within-group longitudinal comparisons relative to baseline (day 1) levels. This approach allows not only to characterize the transcriptional programs that are being regulated for each vaccination group, but it also controls for interindividual differences in pre-existing transcript levels. When compared to baseline levels within each group, G4 vaccinees showed significant induction or downregulation of at day 3 (2321 genes) and 7 after priming (81 genes), while no changes were detected after the prime dose with LAIV or placebo (G1, G2 and G5, respectively) (Figure 2B-C). The longitudinal comparisons also confirmed that after the booster dose, only the subjects who received the IIV5-AS03 (G1) showed significant induction of gene expression (80 genes). Altogether significant transcriptional responses as early as day 3 after vaccination anticipated significant changes in stalk antibody levels measured 29 days later (Figure 1C).
IIV8/AS03 – IIV5/AS03 administration induced early activation of cell signaling and innate immune pathways associated with B cell proliferation and induction of anti-stalk antibodies
To better understand the relationship between anti-stalk antibody induction and the dynamic regulation of gene expression longitudinally, we expanded our previous analysis and investigated the functional pathways perturbed after prime - boost in G4 (IIV8/AS03 – IIV5/AS03) vaccinees. We first analyzed changes early after priming (day 3 and day 7). As shown in Figure 3A, significant differences were mostly found on day 3, with 1317 upregulated genes versus 1004 genes downregulated. The number of upregulated genes decreased on day 7 (80 genes) and only two genes were found to be downregulated. By day 29, no significant differences were found relative to the vaccine regimens suggesting that gene expression changes were back to baseline. We next used this analysis to perform gene ontology (GO) enrichment analysis (Figure 3B) and to build volcano plots (Figure 3C-D). The gene expression changes were associated with 59 (up-) and 33 (down-) biological processes (BP) categories on day 3 (Supplemental Dataset 1), which included innate immune responses and responses to virus infection. Particularly, the top-10 enriched GO categories of induced genes included cytokine signaling and immune cell activation (Figure 3B). Genes included in these pathways were classical genes induced in response to virus exposure such as ISG15, IFIT3, OAS1, DDX58 (RIG-I) or SERPING1, among others (Figure 3C and Table S1). In contrast downregulated BP included chromatin organization and regulation of transcription and histone acetylation, suggesting cell remodeling changes triggered by immune signaling. On day 7, the top-10 GO categories of upregulated genes transitioned from innate immunity to activation of adaptive immune cells (Figure 3B), including leukocyte migration, phagocytosis and recognition, and regulation of B cell activation. In total there was enrichment of 17 GO categories on day 7 post-prime (Table S1), aligned with a high upregulation of immunoglobulin transcripts and the CD38 cell activation marker (Figure 3D and Table S1). When looking at the top upregulated immunoglobulin transcripts, we found that the response was dominated by a variety of clonotypes, with overrepresentation of IGKV1-39 followed by IGHG1 > IGLV1-44 > IGKV1D-39 > IGHG3 (Figure 3D). As expected, by day 29 gene expression reverted to pre-vaccination levels and no significant differences were found when the peak of the antibody response was detected by serological methods. Similarly, no significant changes of the transcriptional profile were detected on day 85, before the booster dose, or on day 93 (post-boost), consistent with a minimal increase of anti-stalk antibodies in this group (Figure 1C).
The immune response is a complex biological process that requires multiple interactions among immune cells with heterogenous differentiation states. We next decided to interrogate the function and relative cell-type composition of the cells circulating in blood after vaccination with the cHA influenza vaccine. For this we used transcriptional signatures obtained from healthy peripheral blood mononuclear cells (PBMCs) profiled by bimodal protein-RNA measurements with Cellular Indexing of Transcriptomes and Epitopes by Sequencing (CITE-Seq) at the single-cell level16. These cell type-specific gene expression signatures consisted of 58 cell types and subtypes, including subsets of the T, B and myeloid cell compartment. To simplify cell type groups, subclusters were collapsed into 37 major cell types. We applied a false discovery rate (FDR) < 0.05 (adjusted p value) to deconvolute the bulk DEG data and determine cell type specific signatures enriched or depleted in the peripheral blood of G4 after priming with IIV8/AS03 on day 3 and 7 (Figure 4A). Fold-change gene signatures revealed a significant enrichment for monocytes, both classical-CD14+ and non-classical-CD16+, and different subsets of dendritic cells (DC) on day 3 post-prime. This was consistent with the biological processes found to be enriched in the GO analysis, as these populations are main mediators of cell signaling processes, innate immune responses and antigen presentation. Enrichments on day 7 suggested that relative cell type composition changed to an enrichment of B cells with different maturity state: memory > intermediate > naïve of the kappa and lambda chain expressing cells. At this timepoint a significant enrichment was also found for antibody-secreting cells: plasma cells > plasmablast confirming the presence of affinity matured long- and short-life antibody producing cells induced by the universal vaccine prototype. Additionally, plasmacytoid DCs (pDCs) were also induced at day 7. This is in contrast to conventional DC (cDCs) that were enriched earlier, on day 3 post-prime. Figure 4B shows the calculated fold-change for 29 cell types, including subsets from the main cell type lineages: CD4+ and CD8+ T cells, B cells, DCs, monocytes, and natural killers (NK) on day 3 and 7 post-prime.
Finally, we quantified the diversity of immunoglobulin (Ig) transcripts that were differentially expressed, as a proxy for the induction of antibodies in response to vaccination. The relative composition of enriched isotypes and subclasses, and the heavy, kappa light and lambda light chain (IGH, IGK and IGL) loci usage for the variable (V) regions on day 7 after priming with IIV8/AS03 are shown in Figures 4C and 4D. Enriched transcripts included members of all loci (Figure 4C). The majority of the response of the constant (C) region was dominated by IGHG and IGHM isotypes, while the most abundant subtypes were, as expected, IGHG1 > IGHG3 > IGHG4 (Figure 4D). For the V regions, the majority of IGH genes were from the IGHV3 (35%) and IGHV4 (32%) subgroups followed by IGHV1 > IGHV5. For the IGK locus, enriched transcripts were dominated by those from the IGKV1 (51%) and IGKV3 (24%) and IGKV4 (23%) subgroups, while the IGL locus was dominated by IGLV1 (36%) followed by IGLV2 (33%) and V3 (24%), and included transcripts from at least 6 different loci 17.
Early and strong induction of the antiviral response correlates with high anti-stalk antibody titers
Next, we asked whether individual differences of the induction of specific antibodies targeting the stalk of the HA of group 1 influenza viruses after vaccination with IIV8/AS03 – IIV5/AS03 (G4) could be linked to specific and unique transcriptomic signatures. First, we looked at individual cH6/1 IgG antibody responses before (pre-) and 29 days (post-) after prime. All vaccinees showed similar antibody titers post-prime (Figure 5A). We next calculated the fold rise on cH6/1 antibodies as the ratio between post- antibody value to pre- levels for each vaccinee. We then computed the geometric mean rise (GMR) by taking the exponent (log10) of the mean fold rise of all individuals. As shown in Figure 5B, GMR (95% CI) was 12.1 (6.9-21.2). We then set a GMR value of 10 as a threshold of higher (>GMR, above the geometric mean) versus lower (≤GMR, below the geometric mean) vaccine responders. Comparison of both low (n = 6) and high (n = 9) responders at baseline showed no pre-existing differences that could explain the variability in the magnitude of the antibody response after vaccination. We next investigated differences on gene expression on day 3 and 7 post-vaccination by comparing each group to their respective baselines (day 0). A total of 469 (390 up and 79 down) and 942 (620 up and 322 down) genes were differentially expressed for the low and high responders, respectively, on day 3. Among these genes, 310 were upregulated in both groups, all of them related to canonical immune response processes (Figure 5C-D). Gene ontology analysis of genes that were exclusively up (n=80) in the low responders returned no significant GO annotated processes, whereas those that were exclusively up in the high responders included additional genes mainly related to innate immune response and cytokine mediated signaling (Figure 5D, top 10 categories). In addition, processes exclusively upregulated in high responders included mitochondrial organization, respiratory chain and upregulation of transcription factor activity (Supplemental Dataset 2). For the downregulated genes, those that were common or exclusive to either group did not return relevant GO categories implicated in immune activation (Supplemental Dataset 2). We then compared levels of the top-10 expressed genes between individuals with low and high antibody titers. Results showed that, while non-significant, expression across the top DEGs was lower in individuals with fold induction of anti-stalk antibodies below the geometric mean (Figure 5E). While differences in group size (low, n=6 vs high n=9) could affect the power to detect DEGs, the differences in absolute expression levels in each group suggest that the overall differences between low and high responders was due to the magnitude of the induction rather than the expression of specific gene programs. On day 7, the transcriptional response for both low- and higher- responders converged in the expression of an adaptive immune response signature, including complement activation, leukocyte migration and B cell activation (Figure 5D). Notably, the levels of expression for Ig genes on day 7 were similar regardless of the magnitude of the measured antibody response (Figure 5F).
G1 and G4 induce similar functional patterns but different immunoglobulin repertoires
As shown before, priming with LAIV8 did not induce specific antibodies against the stalk of the HA protein (Figure 1C). However, LAIV-primed individuals showed a significant increase of antibody titers after boosting with IIV5/AS03, or IIV5 only, in the G1 and G2 groups, respectively. We aimed to characterize the functional pathways perturbed by the booster regimen. We interrogated G1 first since antibody responses on day 113 (29 post-boost) were higher compared to the other vaccination groups after the booster dose. Longitudinal changes of gene expression were determined after normalization by baseline or day 85 (pre-boost) as indicated in Figure 6A: day 85, 92 and 113 using baseline as a reference; or day 92 and 113 using day 85 as a reference. No significant differences were found between baseline before priming (day 1) and pre-boost (day 85), and DEGs at day 92 showed comparable transcriptomic profiles independent of the reference used. Enrichment of 72 genes related to B cell activation and immunoglobulin transcription were identified: IGHG1 > IGLV1-44 > IGKV2D-29 > IGHV1-69 > IGKV2-28. Remarkably, we found preferential usage of IGHV1-69 among the top-10 DEGs. Additionally, IGHV1-69-2, and IGVH1-18 were also found among the top-25 genes induced after the boost. Since anti-HA stalk antibodies are preferentially encoded by immunoglobulin heavy chain V region gene VH1-69 and VH1-1818, usage of these clonotypes indicated that the vaccine used in this trial induced specific and not generic immune responses against the stalk of group 1 HA protein. Intersection of the top-10 most significant biological processes, and top-10 significant genes at day 92 are shown in Figure 6B-C. Finally, gene expression levels at day 113 (day 29 post-boost) were restored to baseline levels before vaccination, similar to those on day 29 after prime.
Before we showed that G1 (LAIV8-IIV5/AS03) and G4 (IIV8-IIV5/AS03) induced similar antibody levels and functional patterns after receiving the inactivated adjuvanted cHA vaccine for the first time, either after boosting or priming, respectively (Figure 2A). Since G1 vaccinees were primed with LAIV8, we asked whether relative cell type composition and relative Ig repertoire would also be the same. Cell type deconvolution (shown in Figure 6D) indicated similar enrichment of cell types enriched in G1 after boost compared to G4 after prime (Figure 4A). However, some differences were found when comparing the Ig transcripts encoding for the C region of immunoglobulin heavy chains. Ig alpha (IgA) is the major immunoglobulin class in body secretions and can be found on linings of the respiratory tract, digestive system, and saliva19. It was expected that the administration of LAIV8 would induce mucosal antibodies at the mucosal surfaces. However, saliva and serum samples analyzed by enzyme-linked immunosorbent assays (ELISAs) for anti-stalk IgA and secretory IgA (sIgA) showed no such antibodies in G1 or G2 after LAIV administration6. Contrary to this, when we look at the relative abundance of the Ig repertoire transcripts (Figure 6E-F), we found that it was still dominated by IGHG, but vaccinees from the G1 showed also an enrichment of the IGHA subclass (19%). This contrasted with G4, in which the response was dominated by IGHG followed by IGHM and no IGHA transcripts were detected (Figure 4C). Except for the presence of IGHA1 transcripts, the composition of transcripts from the C region in G1 (Figure 6F) was similar to G4 (Figure 4D). However, the diversity and the relative proportion of Ig transcripts at the V chains from the IGH, IGK, and IGL loci was also different from those elicited by the G4. Subgroups of encoded transcripts for G1 included IGHV3 (46% each), IGHV1 (35%), and IGHV4 (19%) only, whereas the IGK responses were dominated by the IGKV3 subgroup (IGKV3 and IGKV3D accounted for 45% of IGKV detected transcripts) followed by the IGKV3 (29%) and IGKV2 (6%). Lastly, the induced IGL transcripts were less diverse compared to G4, with only IGLV1 (57%), IGLV3 (23%), and IGLV2 (20%) subgroups being expressed. The immunoglobulin repertoire on day 7 after boost with IIV5/AS03 across the differentiation spectrum is shown in Figure 6E. The observed differences in the repertoire of Ig transcripts and detection of IgA expression shows that the intramuscular adjuvanted vaccine effectively boosted previously undetectable mucosal immunity induced by the LAIV8 prime. This in contrast with G2 for which the only difference in vaccine regime was the presence of adjuvant in the booster dose.
Finally, similar to G4, we looked at expression levels across individuals that developed low (≤GMR) or high (>GMR) cH6/1 antibody titters, as measured by ELISA. Both groups showed similar levels of gene expression on day 7 post-boost, regardless of measured antibody titers at the late timepoint post-boost, on day 113 (Figure 6G). This was independent of pre-existing immunity as Ig transcript levels before priming or boosting showed no correlation with induction of antibodies levels.
Differences of vaccine induced antibody responses attributed to the adjuvant
Because we observed that higher IgG responses in serum were associated with the use of the AS03 adjuvant after boosting with the inactivated formulation, we interrogated the specific contribution of the adjuvant to the gene expression signatures detected. For this, we compared DEGs between G1 (IIV5/AS03) versus G2 (IIV5) on day 7 post-boost in parallel to DEGs between G1 (IIV5/AS03) and G5 (PBS) at the same time point. Analysis showed 93 (up-) and 28 (down-) genes for the G1 vs. G2 comparison; and 29 (up-) and 10 (down-) for the G1 vs. placebo (Figure 7A-B). When we compared the GO categories enriched, we determined that both comparisons shared categories such as complement activation (classical pathway), leukocyte migration, and regulation of B cell activation. In contrast, some categories were unique for the G1 vs. G2 including phagocytosis (recognition and engulfment), immune-activating cell surface receptor signaling pathway or protein folding (Figure 7C). Finally, we considered that the comparison of G1 versus G2 assessed the contribution of the adjuvant independently (although in the context) of IIV5, while the comparison of G1 versus G5 – placebo – assessed the contribution of both IIV5 and AS03. If we then intersect the datasets obtained, the induction or depletion of those genes in common should be driven by the use of the AS03 adjuvant independently. As shown in the Venn-diagram (in Figure 7D), 27 of the induced genes were shared between both datasets. While 12 of them were directly related to transcription of Ig: IGHG1, IGHG3, IGKV1D-39, or JCHAIN among others; 15 referred to other processes such as regulation of innate and adaptive immune responses, including complement activation, response to stress, and B cell development: IRF4, XBP1, ITM2C, MZB1, PDIA4, POU2AF1, STT3A, TNFRSF17 (Supplemental Table 1).