3.1 Breadth and magnitude of IgG response is influenced by TLR agonist(s) and emulsion. Groups of C57Bl/6 female mice (7-10 weeks old) were administered via s.c. route (base of tail) a single dose of trimerized VN04 H5 in a different TLR agonist (TLRa), or combination TLRa, and formulated without or with squalene oil-in-water emulsion (AddaVAX) as shown in Table 1. Agonists comprised Pam2CSK4 (TLR2/6a), pyrimido-indole derivative and monophosphoryl lipid A (both TLR4a), imidazoquinoline derivative (TLR7a), ODN 1018 CpG DNA containing a phosphorothioate backbone (TLR9a), and a chemically linked tri-agonist (TLR2/6_4_7) as described previously (23). The pyrimido-indole TLR4a is less potent than the conventional MPLA agonist for TLR4, but is more suited for tri-agonist synthesis (23, 39). Plasma samples were obtained at days 10, 28, 43 and 103 for probing against an influenza protein microarray described previously (25), which displayed 125 different monomeric full length HA (HA0) drift variants and 131 HA1 variants, spanning all 18 HA subtypes. Other than transient weight loss in the first 24h after vaccination, the animals in each group showed normal increases in body weight (Supplementary Fig. S1).
IgG responses peaked ~d28-d43; the IgG profiles by protein microarray for full-length HA0 and HA1 on d28 are shown in Figs. 1 and Fig. 2, respectively. Full-length H5 drift variants on the array range overall from 90.3 - 99.8 (average 96.4) % sequence identity with the immunizing HA (VN04 H5), while drift variants of H1 (the closest phylogenetic group to H5) range from 59.5 - 65.6 (average 62.9) % sequence identity with VN04 H5. These values are higher if the conserved HA2 regions are compared. Thus, 92.9 – 99.6 (average 97.8) % identity for H5 variants, and 75.7 – 77.9 (average 77.1) percent identity for H1 variants. In addition to H1, H5 immunization also induces heterosubtypic cross-reactivity for other closely-related Group 1 HAs, including H2 and H6. In the absence of AddaVAX emulsion (orange symbols in Figs. 1 and 2), the strongest responses were obtained with MPLA and the CpG/MPLA combination adjuvant (Figs. 1G and H). The linked TLR 2/6_4_7 tri-agonist and TLR 2/6a also induced robust H5 responses (Figs. 1A and C), while unlinked TLR2/6,4,7 tri-agonist, TLR7a, and TLR9a induced weak, but measurable homosubtypic IgG cross-reactivity only without emulsion (Figs. 1B, E and F, respectively). H5 with pyrimido-indole derivative (TLR4a), or administered alone without TLRa, failed to induce detectable IgG under the conditions used (Figs. 1D and I, respectively). AddaVAX alone was an excellent adjuvant for H5 (Fig. 1I), and consistently elevated the responses induced by different TLR agonists. The largest effects on antibody responses by addition of AddaVAX (AddaVAX deltas) were seen when the response to TLR agonists alone was low, namely the unlinked TLR2/6,4,7 tri-agonist (Fig. 1B) and TLR4, TLR7, and TLR9 agonists (Figs. 1D, E and F, respectively). Of particular note, the pyrimido-indole derivative (TLR4 a), which failed to elicit detectable IgG against H5, produced robust homo- and hetero-subtypic cross-reactivity when co-administered with AddaVAX. This was higher than achieved with AddaVAX or TLR4a alone, indicative of a synergistic response. AddaVAX deltas were smaller where the TLR agonists alone induced responses, i.e., linked TLR2/6_4_7 tri-agonist (Fig. 1A), and TLR2/6, TLR4 and TLR9/4 agonists (Figs. 1C, G and H, respectively). Nevertheless, the effect of the emulsion was often dramatic. For example, the cross-reactive response to H5 administered with unlinked tri-agonist, TLR7a, or TLR9a, were modest and homosubtypic only, whereas when the same agonists were co-administered in AddaVAX induced near maximal homosubtypic and heterosubtypic signals. This enhancement of heterosubtypic cross-reactivity by AddaVAX could potentially increase the utility of the vaccine to provide protection against novel subtypes with pandemic potential. Interestingly, the smallest AddaVAX delta was seen with MPLA (a TLR4a), which is by itself a strong adjuvant for H5; the profile after coadministration with AddaVAX was essentially unchanged. However, this may be related to the liposome needed for delivery of hydrophobic MPLA which may not benefit from an oil-in-water emulsion in the same way as other water-soluble agonists.
In addition to full-length HA, the protein microarray also displays HA1 fragments, which contain the variable head domain and part of the more conserved stalk of the protein. Previous studies with this array format have shown that sera with reactivity to the full-length (HA0) may, or may not, also recognize HA1, depending on whether there are antibodies present against the head (25, 40). IgG profiles to HA1 fragments are shown in Fig. 2. The TLR9a/4a (CpG/MPLA) combination adjuvant in AddaVAX was the only formulation able to engender robust IgG homosubtypic cross-reactivity for HA1 fragments of H5 (Fig. 2H). Modest responses were also generated using the individual TLR9a and TLR4a (MPLA) agonists alone when administered with emulsion (Figs. 2F and G) although the IgG signals when combined (Fig. 2H) were greater than the sum when used alone. Of the other adjuvants tested, only the unlinked combination of TLR2/6a, TLR4a and TLR7a in the presence of emulsion (Fig. 2B) were able to induce IgG. The weaker response to HA1 compared to full length HA0 may reflect the smaller size of the HA1 polypeptide, and consequently the fewer B cell epitopes available for recognition by the polyclonal response. Alternatively, the conformation of HA1 may not accurately reflect the conformation of the same sequence within the trimerized full-length protein, which may also contribute to the reduced strength of the response. To this point, we presented data previously (25) showing that conformation-sensitive mAbs raised against H1 from the 2009 pandemic recognized both HA0 and HA1 on the array, suggesting the HA1 is folded correctly. Combination adjuvant CpG/MPLA/AddaVAX was the only adjuvant to engender detectable hetero-subtypic cross-reactivity, albeit against only one H1 and one H2 variant (Fig. 2H). None of the other adjuvants tested were able to stimulate heterosubtypic cross-reactivity for HA1 fragments.
H5 has undergone antigenic drift in poultry and natural wildfowl populations, and different H5 variants are typically grouped into clades and subclades (41). A successful pre-pandemic vaccine will need to provide broad coverage across different clades to provide anticipatory protection. Thus, Fig. 3 shows array data of H5 variants by clade, which reveals broad cross-clade reactivity generated by the majority of adjuvants tested. In the absence of AddaVAX, cross-clade reactivity is the strongest in TLR9a/TLR4a, (CpG/MPLA), TLR 4a (MPLA), linked TLR 2/6, 4_7a, and TLR 2/6 adjuvant groups. In the presence of Addavax, cross-clade reactivity is elevated for all adjuvant groups.
In addition to antibody profiles, we assessed the dynamics of the response. Fig. 4 shows the development of homosubtypic reactivity at 3 time points after a single dose. In the absence of AddaVAX, the TLR9aTLR4a (CpG/MPLA) combination adjuvant produced the most rapid response, being the only vaccine to have elicited detectable antibodies on day 10. By day 28, IgG reactivity was also seen from linked TLR 2/6_4_7a tri-agonist, TLR2/6a, and TLR4a (MPLA), which remained elevated at day 42. In the presence of AddaVAX, all the vaccines elicited reactivity by day 28, although CpG/MPLA/AddaVAX again elicited the most rapid response, as can be seen on day 10. We conclude from these data, and data presented in Fig. 1 and 2, that of all the adjuvants tested, CpG/MPLA/AddaVAX induces the most rapid response, with the greatest magnitude and broadest homo- and hetero-subtypic cross-reactivity.
3.2 Th1/Th2 balance is variable according to adjuvant used. Both IgG1 v IgG2c subtyping on microarrays and cytokine profiling in T cell recall assays were performed on the same groups of immunized C57Bl/6 mice to evaluate the Th1/Th2 response elicited by each adjuvant tested (Fig. 5). In the absence of AddaVAX emulsion, TLR2/6_4_7a tri-agonist, and individual TLR2/6a and TLR7a stimulated IgG1 antibodies (Th2) (Fig. 5A), while TLR9a was strongly polarized to IgG2c (Th1), and the combination adjuvant TLR9a/TLR4a (CpG/MPLA) gave a more balanced response. Overall, the addition of AddaVAX elevated the IgG1 and IgG2c signals but did not reverse any polarization induced by the TLR agonists alone. Thus, AddaVAX elevated the IgG1 signals to both linked and unlinked 2/6_4_7 tri-agonists, as well as individual TLR 2/6, 4 and 7 agonists. Responses to TLR9a (CpG) and TLR4a (MPLA) and the combination TLR9/4a (CpG/MPLA) were more balanced with both IgG1 and IgG2c detected. It is noteworthy that in the absence of AddaVAX responses to some TLR agonists were undetectable (i.e., unlinked tri-agonist, and individual TLR4 (PID) and TLR9 agonists); however, in the presence of AddaVAX all three were strongly immunogenic. Finally, we noted H5 antigen with AddaVAX alone (i.e., without TLR agonists) was also balanced.
IFN-g and IL-4 release was determined from the same mice at the experimental endpoint by ELISpots (Fig. 5B). The recall assay included immunizing (H5) antigen, as well as hemagglutinins H1 and H7 (63% and 42% sequence identity with H5, respectively). Despite observing heterosubtypic cross-reactivity at the Ab level, IFN-g and IL-4 spot-forming cells (SFCs) were detected against H5 only. Several cytokine patterns were consistent with IgG1/IgG2c profiles. For example, in the absence of AddaVAX, the only two groups to induce IgG2c were agonists to TLR4 (MPLA) and the combination TLR9/4a (CpG + MPLA) adjuvant; accordingly, these also produced H5-specific IFN-g SFCs in the recall assay. In the presence of AddaVAX, these two agonists elicited elevated signals to IgG1 which correlated with the presence of IL-4 SFCs. Similarly, in the absence of AddaVAX, linked TLR 2/6_4_7 tri-agonist and TLR 2/6 and 7 agonists produced only IgG1 (Th2) and no IFN-g SFCs, of which two (linked TLR tri-agonist and TLR 2/6 agonist) produced detectable IL-4 SFCs. In the presence of AddaVAX, linked and unlinked TLR tri-agonist, and individual TLR 2/6, 4 and 7 agonists produced strongly polarized IgG1 (Th2) responses and correspondingly low numbers of IFN-g SFCs but detectable IL-4. This contrasts with AddaVAX alone, which produces a balanced IgG1/IgG2c profile, suggesting these TLR agonists suppress the production of IgG2c. Of note is the TLR4a, PID, which appears ineffective at inducing antibody or cytokines in the absence of AddaVAX, but which becomes a strong Th2-polarizing adjuvant in the presence of AddaVAX, displaying strong IgG1 signals and robust IL-4 SFC numbers. Note this contrasts with the combination TLR 9/4a (CpG + MPLA) in the absence of emulsion, which strongly polarizes the response to IgG2c (Th1) and a production of IFN-g SFCs. Thus, the breadth of the cross-reactive response and polarization can be tuned according to the adjuvant, as summarized in Table 2.
The largest numbers of IFN-g and IL-4 producing SFCs was induced by the combination of MPLA, CpG and AddaVAX. This appears to result from synergy between AddaVAX and either TLR agonist, as the SFC numbers for MPLA+AddaVAX or CpG+AddaVAX were greater than the sum of SFC numbers for the individual components. Measurements of antigen-specific B cells also suggest these components are synergistic (see section 3.3 below).
Additional cytokines were assayed in the supernatants of recall assay using multiplex bead detection (Fig. 5C). The highest concentrations were elicited by MPLA/AddaVAX and CpG/MPLA/AddaVAX combination adjuvants. The former produced high levels of Th1-associated IL-2 and TNF-a, as well as high levels of IL-6, which is produced by different APCs and promotes Th2 differentiation (42). These data are consistent with the balanced IgG1/IgG2c profile seen by arrays and ELISPOTs. CpG/MPLA/AddaVAX also produced high levels of IL-2 and TNF-a, but lower levels of IL-6, which may contribute to the skewing toward IgG2c over IgG1. Both MPLA/AddaVAX and CpG/MPLA/AddaVAX also induced modest levels of Th17-associated cytokines. Among the non-emulsified formulations (left side of Fig. 5), of note is CpG/MPLA. This adjuvant is strongly biased to Th1 with elevated IL-2 and TNF-a but low IL-5 and IL-6, consistent with the strongly polarized IgG2c response. MPLA produces a more balanced Th1/Th2 cytokine profile, consistent with the balanced IgG1/IgG2c response. The remaining non-emulsified agonists, which produced mainly IgG1 skewed responses, produced both Th1 and Th2-associated cytokines. Interestingly, H5 alone (without TLRa or emulsion), which did not induce detectable antibodies, was associated with the release of several cytokines in the recall assay. This profile was amplified by AddaVAX, notably release of IL-5. This would suggest H5 has an inherent capacity to stimulate a T cell response, as has been reported by others (43, 44), although the mechanism is unknown.
3.3 CpG and MPLA synergize in the induction of antigen-specific plasma cells. It was of interest to determine whether the effects of combining CpG and MPLA in AddaVAX were synergistic, since synergies may allow lower doses of adjuvant to be used and development of less reactogenic vaccines. To address this, we used antigen-specific B cell labelling to quantify these in mice administered influenza HA in CpG/MPLA/AddaVAX, and compared these with CpG/AddaVAX or MPLA/AddaVAX individually. The CpG/MPLA/AddaVAX combination adjuvant induced more total plasma cells (Fig. 6A) and antigen-specific B cells (Fig. 6B) than the sum of CpG/AddaVAX and MPLA/AddaVAX separately, indicating a synergistic effect. Almost all the antigen-specific B cells were CD138+, indicating their differentiation into plasma cells. These numbers are shown as bar charts in Fig 6C. IgG isotype staining of antigen-specific B cells revealed both IgG1 and IgG2a are produced (Fig. 6D), although the majority expressed IgG2a, consistent with the Th1>Th2 polarization seen by array profiling, and T cell recall assay (Fig. 5).
3.4 Further characterization of the utility of IVAX-1 as an adjuvant. Overall, the data presented thus far show that CpG, MPLA and AddaVAX, a combination adjuvant we term ‘IVAX-1’, synergize to induce the most rapid and broadest cross-reactive response among all the adjuvants tested. Therefore, we performed a series of experiments to characterize IVAX-1 further. Fig. 7A shows plasma generated using VN04 H5/IVAX-1 neutralized H5N1 reassortant virus in vitro. We noticed, as reported previously (25), that boosting was required to reach detectable levels of neutralization. We showed earlier that IVAX-1 engendered a robust IFNg-response in the ELISpot (Fig. 5B). As the antigen for recall was supplied as soluble H5, it was expected that most of the processed antigen would be presented in the context of class II MHC to CD4 T cells, rather than to CD8 T cells. This was confirmed when intracellular cytokine staining (ICS) was used in place of the ELISpot to identify the source of IFNg (Fig. 7B). Therefore, we attempted to improve the recall of CD8 T cells using H5N1 influenza virus-infected splenocytes (vAPCs). However, we saw no improvement in detection of antigen-specific CD8 cells. Although the data support the notion that IVAX-1 preferentially stimulates CD4 T cells, we cannot exclude a CD8 T cell response as infected respiratory epithelial cells (the natural host cell for IAV) may be required for vAPCs instead of splenocytes. Finally, we explored intranasal (i.n.) administration of IVAX-1, since the above experiments were based on the s.c. route only. The i.n. route is often used to engender an IgA response in the nasal and respiratory mucosa. Mice were administered H5/IVAX-1 via the i.n. route and boosted on d56. Other than transient weight loss (<8%) within 24h of the prime and boost, there were no adverse events associated with this intranasal delivery or boosting (not shown). At the experimental endpoint (d69), IgG, IgG1, IgG2c and IgA profiles were determined for plasma, and nasal and lung washed by microarray (Fig. 7C). Robust IgG responses were seen in plasma, dominated by subtype IgG2c, consistent with the findings seen when the same vaccine was administered by the s.c. route, as shown in Fig. 5. Robust IgA responses were found in nasal and lung washes, and IgG was also found in the lung, although no IgA was detected in plasma. Ab profiles recognized by IgG, IgG2c and IgA were identical after i.n. administration, and comprised recognition of both H5 (homo-subtypic) and H1 (hetero-subtypic) variants. Indeed, the IgG profile induced by i.n. administration was highly correlated with that induced by s.c. administration (Fig. 7D).
3.5 Single-cell mRNA sequencing (sc mRNA seq) analysis. Finally, in order to generate a transcriptomic “fingerprint” of IVAX-1 for future comparison with other adjuvants, we performed an unbiased single cell transcriptomic analysis of whole spleen cell suspensions pooled from 3 immunized B6 mice that received H5/IVAX-1 barcoded using the 10x Genomics platform and sequenced on the Illumina HiSeq, and performed a similar analysis on an aged-matched naïve control B6 mouse for comparison (see methods). To our knowledge, this is the first such dataset generated from whole lymphoid tissue without prior cell sorting. A previously trained supervised support vector machine model (SVM) was first used to define different cell types based on transcriptional profiles. The majority of splenocytes were of lymphoid origin with some myeloid cells, as shown in the Uniform Manifold Approximation and Projection (UMAP) plot in Fig. 8A and Supplemental Fig S2A. The heatmap in Fig. 8B shows expression profiles of transcriptionally active splenocytes in mice receiving H5/IVAX-1. Cell populations were identified by differential gene expression of the subtype compared to all other subtypes. Overall, three main arms are delineated: T lymphocytes (which also includes NK/T cells, gd T cells, and Tregs), B lymphocytes, and myeloid cells (myeloid dendritic cells, macrophages, monocytes and neutrophils). Interestingly, the unbiased hierarchical clustering of cell type transcription profiles placed B cells more closely related to myeloid cells than to T cells. The major cell populations were well delineated by marker genes. Thus, the T cells cluster show elevated expression of Trac, Trbc2, Cd3g and Cd3d (TCR a and b chain constant domains, and CD3g and d chains, respectively). Others include Ms4a4a and b (CD20 homolog expressed on mature Th1 cells (45)); Tcf7 (T cell-specific transcription factor (46)); and Skap1 (regulator of TCR signaling through MAP kinase, (47)). Nkg7 and Ctsw genes (NK cell granule protein 7 and cathepsin W) are particularly elevated in NK/T cells and play roles in cell-mediated killing by NK and CD8 T cells (48, 49). Also of note is expression of Il2rb (IL2 receptor b chain) which is expressed in NK/T and regulatory T cells (Tregs) at this time point. Also expressed in Tregs is the signature gene Izumo1r (Folr4; folate receptor) (50, 51). The B cells arm shows elevated expression of Cd79a and Cd79b (BCR CD79a and b chains); Ebf1 (B cell-specific transcription factor (52)); Ighm (IgM heavy chain); and Cd74 (MHC class II invariant chain (53)). Cells in the myeloid cell express Fcer1g (IgE Fc receptor); Tyrobp (an immunoreceptor-associated tyroskine kinase); and Ifitm3 (IFN-induced transmembrane protein 3), an anti-viral protein that which restricts the entry of several viruses into host cells (54). Among the myeloid cells, neutrophils show expression of S100 family proteins, including S100a8 and S100a9 (myeloid-related proteins 8 and 14), which are known to be released by activated phagocytes and amplify the inflammatory effect of LPS mediated through TLR4 (55). Also expressed in neutrophils is Msrb1 (methionine reductase) which appears to enhance the expression of anti-inflammatory cytokines by macrophages (56), and Hp (haptoglobulin), a major acute phase protein. These data serve as an additional verification that the SVM correctly identified the cell types. Interestingly we found a small population of B and T cell doublets which were identified due to their expression of both T cell markers (Cd3g) and B cell markers (Cd79a) and an unbiased doublet detection algorithm (Supplementary Fig. S2B-D). Doublets in droplet-based sequencing platforms can be indicative of juxtacrine signaling.
We then asked whether the differential IgG2c>IgG1 response characterized above could be corroborated using the transcriptomic data. We found Ighg2c, Ighg2b and Ighg1 genes were all upregulated in B cells from immunized mice relative to naïve mice (Supplementary Fig. S3 and Fig. 8C), with log2-fold changes of 6.21 (p-value = 2.14*10-93), 4.35 (p-value = 2.33*10-281) and 2.69 (p-value = 1.10*10-158), respectively (Mann-Whitney U rank-sum test). This is consistent with the moderately skewed IgG2c>IgG1 response observed by IgG isotyping after IVAX-1 administration (Fig. 5a). We also compared the expression of the Ifng gene in naïve and immunized mice (Fig. 8D), which showed IFNg was expressed by both CD4 and CD8 T cells, with log2-fold increases of 0.7 (p-value=0.28) and 0.9 (p-value=0.69), respectively (Mann-Whitney U rank-sum test), consistent with a robust IFNg response in T cell recall assays (Fig. 5B). Interestingly, intracellular cytokine staining suggested IVAX-1 was not a strong inducer of antigen-specific CD8 cells (Fig. 7B), although as mentioned above, this may reflect suboptimal restimulation conditions for CD8 cells in the recall assay. Overall, these data show that the transcriptomic Th1/Th2 balance matched that defined by IgG subtyping and cytokine expression in T cell recall assays.
We also inferred cell-cell signaling interactions between different cell types using CellChat, a recently developed tool that defines intercellular signaling strengths between specified cell types in single-cell transcriptomic data. This tool has previously been used to analyze differential wound healing signaling in adult mice (38). As shown in Fig. 8E, lymphocytes (i.e., NK/T cells, T cells, Tregs and B cells) communicate predominantly with macrophages before immunization, whereas this shifts to dendritic cells and neutrophils after immunization. The gd T cells, B/T cell doublets and resident monocytes were excluded from the analysis because they did not meet the cutoff of 100 cells. Intra-cell type communication (e.g., T cells to T cells) is higher for lymphocytes in unimmunized mice and higher for neutrophiles and dendritic cells in immunized mice. Overall, these data provide a useful first look at the transcriptomic and cellular interactome induced by IVAX-1 at a single time point. Similar studies at different time points after vaccination, and with different adjuvants for comparison are in progress.