The vaccination and sampling regimen is summarised in Fig. 1 to provide an overview of the study design and experimental setup.

Prime and boost vaccines were delivered via a recombinant (r) modified vaccinia virus Ankara (MVA) vector and in form of an adjuvanted water-in-oil-in water (W/O/W) protein subunit formulation, respectively. a In stage I (n = 1), elephant A was vaccinated with 1 × 10⁸ plaque-forming units (PFU) of rMVA-EE2-MCP and 25 µg of rEE2 and rMCP antigen (Ag) each in a first prime-boost cycle (cycle 1). In a second prime-boost course (cycle 2), the animal received 5 × 10⁸ PFU of the rMVA vaccine and 50 µg of each recombinant vaccine Ag in W/O/W. All vaccinations in stage I were administered at 3-week intervals. b In stage II (n = 2), elephants B and C received two prime-boost courses of the higher doses tested for safety in cycle 2 of stage I and vaccinations were set further apart, with 4 weeks between prime and boost vaccinations and 6 weeks between cycle 1 and cycle 2. During both stages, blood and plasma samples were collected as indicated by the red drops below the timelines. Shaded areas indicate the three phases of the sampling and vaccination scheme: grey—pre-vaccination phase; blue—prime-boost cycle 1; red—prime-boost cycle 2. One element in this figure was created with BioRender.com. Elephant photos courtesy of Chester Zoo, used with permission. © Chester Zoo.

Vaccine safety and reactogenicity

Close monitoring of vaccinated elephants raised no safety concerns associated with the recombinant (r) MVA-EE2-MCP prime or the Montanide ISA 201 VG-adjuvanted subunit boost vaccine. No mild or moderate adverse events, for example symptoms such as local swelling, were observed at any point in any of the three elephants vaccinated during this study.

Antigen-specific IFNG gene expression

IFNG as a strong correlate of Th1 immune responses was measured here by RT-qPCR gene expression analysis to assess cellular immune responses following whole blood stimulation with rEE2 and rMCP antigens before and after vaccination (Fig. 2).

The longitudinal IFNG gene expression levels in samples stimulated with rEE2 or rMCP relative to the expression in time-matched phosphate buffered saline (PBS)-treated controls over the stage I (a, b) and stage II (c–f) study periods are shown for the three participating elephants (N = 3). Data are presented for each individual animal (stage I: n = 1; stage II: n = 2). Vertical dashed lines indicate time of vaccination with either rMVA-EE2-MCP virus vector (rMVA) or adjuvanted rEE2-rMCP-subunit (rAg) formulation. Horizontal dashed lines mark the highest IFNG expression level observed among baseline (pre-vaccination) samples. Through calculating the fold changes using the 2^−(ΔΔCt) method, IFNG mRNA levels were normalised to those of the reference gene elongation factor−1α and PBS controls were set as 1. Asterisks below the trial week numbers indicate that a blood sample could not be obtained from an animal at this time point (elephant C: weeks 3 and 16) or samples that were excluded from further analysis due to unsatisfactory downstream quality checks (elephant B: weeks 13, 16 and 20). Source data are provided as a Source Data file.

Latently infected elephants display a background immune reaction against EEHV antigens

IFNG mRNA was detected at significant levels in pre-vaccination blood samples stimulated with either the rEE2 or rMCP vaccine antigen. These recall responses before vaccination are supported by the notion that all three adult elephants were latently infected with EEHV as anticipated for adult animals, which is corroborated by historical EEHV shedding and serological data indicating humoral immunity to major viral glycoproteins⁴².

Pre-vaccination blood samples from all three elephants displayed higher recall IFNG expression levels when stimulated with rEE2 than with rMCP. Interestingly, the baseline IFNG transcript levels correlated to the age of the animals, i.e., the oldest elephant had the highest baseline levels in response to both antigens (Fig. 2e, f).

Vaccination with rEE2 and rMCP antigens induces memory recall responses

The first vaccinated elephant (A) received 20% of the eventual viral dose of MVA and half of the µg protein boost dose during the first rMVA-EE2-MCP prime and protein boost cycle to initially assess safety and determine the dose range, which was then increased in the second prime-boost cycle to the final dose used in the rest of the trial.

Nevertheless, memory recall responses were observed following the vaccination of elephant A: a 948- and 928-fold increase in IFNG gene expression was detected in response to stimulation with the rEE2 and rMCP antigens in week (W) 3 after the first rMVA prime vaccination for this elephant. IFNG transcript levels in rMCP-stimulated blood also started to increase again after the first protein vaccination and peaked at a fold change (FC) of 1178 in W7. Boosting with a second protein injection did not enhance this response. IFNG expression was not boosted above baseline levels in response to the rEE2 antigen during the second vaccination cycle. This suggested that the 3-week intervals between each vaccination did not allow enough time for adequate contraction of IFNG expressing T cells (Fig. 2a, b).

Based on these results, intervals for stage II including elephants B and C were extended to 4 weeks between prime-boost vaccinations and 6 weeks between first and second vaccination cycles (Fig. 1). Both animals received the higher doses from the beginning, and the blood sampling schedule was changed to collecting blood for up to 3 weeks and 1 week prior to vaccination, on vaccination day, and 2, 3 and 4 weeks following each injection to better resolve the vaccine responses over time.

Although substantial variation between elephant B and C occurred regarding the magnitude of recall responses to rEE2 or rMCP antigens in post-vaccination samples, elevated IFNG expression or levels that exceeded baseline values after rMVA or protein boost vaccinations were observed in blood from both elephants after stimulation with rMCP (Fig. 2d, f). In blood obtained from elephant B after initial rMVA vaccination, IFNG recall responses induced upon stimulation with rMCP appeared to be boosted by the first protein vaccination with a 9340-fold increase of IFNG expression in W6 (Fig. 2d). While this was the maximum recall response to rMCP observed for this elephant and downregulation to almost baseline levels supervened, transcript levels were upregulated again and, during the second prime-boost cycle, reached a peak with a FC of 5658 two weeks after the second rMVA vaccination (W12). Although the final protein boost vaccination did not enhance IFNG gene expression in response to rMCP stimulation, IFNG expression remained elevated and above baseline until the end of the study in this animal.

In blood from elephant C, IFNG gene expression was highly upregulated upon rMCP stimulation in W2 and W4 following the first vaccination with rMVA (FC W2: 33,627 and FC W4: 140,572 vs. FC W0: 428; Fig. 2f). While recall responses to this antigen did not exceed W4 levels at any of the following blood sampling dates, IFNG transcript levels were upregulated in response to the remaining vaccinations and appeared to be maintained above baseline following the final protein boost vaccination in this animal (Fig. 2f).

rEE2 stimulation of post-vaccination blood samples induced more modest memory recall responses in elephant B, where IFNG expression only exceeded the highest transcript levels measured in pre-vaccination samples following immunisation with the second rMVA vaccine in this animal (maximum FC: 6767, W14; Fig. 2c). In contrast, IFNG mRNA expression in blood from elephant C was upregulated beyond the highest baseline expression levels following stimulation after the first rMVA vaccination and reached a maximum FC of 108,873 4 weeks after the first protein boost vaccination (W8; Fig. 2e).

Analysis of pre- and post-vaccination whole blood transcriptomes

As IFNG gene expression analysis yielded some encouraging results but retained ambiguity due to the pre-existing immunity to latent infection, we further investigated whole transcriptome changes in the two stage II elephants that had received the full vaccine doses at extended intervals.

Post-vaccination transcriptomic changes in unstimulated blood samples

Our first interest was to identify changes in the transcriptomic profile of control (phosphate buffered saline (PBS)-treated) samples between pre- and post-vaccination time points (contrast A, Supplementary Fig. 1 and Supplementary Data 1a). This revealed that the expression of a total of 2678 genes was significantly modulated after vaccination. Of these differentially expressed genes (DEG), 944 were upregulated and 1734 downregulated. FC of upregulated DEG ranged from 1.3 to 15.6 and from −1.3 to −60.8 in downregulated genes (adjusted p value (padj) ≤0.05).

CD8+ and CD4+ T cells are upregulated following vaccination

To further elucidate the mechanism of action induced by vaccination, we looked for enriched cell signatures associated with the DEG post-vaccination in unstimulated blood. Using the significantly upregulated DEG, over-representation analysis (ORA) identified an association with gene sets specific for CD8+ T cells (p p = 0.003) in the Human Gene Atlas (Fig. 3). The DEG overlapped with 50/602 genes in the Human Gene Atlas CD8+ T cells gene set. Twenty-nine of these genes were shared with the CD4+ T cells gene list, while another 11 genes were specific for CD4+ T cells, thus mapping a total of 40/533 to these.

Over-representation analysis was performed using DEG identified from samples of two elephants (n = 2). A total of 944 upregulated and 1734 downregulated DEG (adjusted p value ≤0.05) were analysed using the Human Gene Atlas database. The red and blue bars represent the combined scores of gene sets enriched for upregulated and downregulated cell types, respectively. In unstimulated blood samples, CD8+ and CD4+ T cells were significantly over-represented, while monocyte and B cell signatures were downregulated following vaccination. Significantly enriched pathways were identified using a one-sided Fisher’s exact test (H₀: no association), with correction for multiple comparisons by applying the Benjamini-Hochberg method, and ranked by the combined score (z-score of the observed and expected means of the DEG and pathways overlap). Pathways with a p value ≤0.05 were considered significant.

Besides the association with the T cell gene sets above, a number of obvious immune mediators, such as interleukin 15 (IL15), CD28 and the inducible T cell costimulator ICOS, were identified among the upregulated genes post-vaccination. Of note, the second most upregulated gene with an FC of 11.5 in the entire gene set was a LOC gene (LOC126061467) predicted to translate into a granzyme A (GZMA)-like protein (Supplementary Data 1a), granzyme A being involved in cytotoxic T lymphocyte-induced apoptosis.

CD14+ monocytes are downregulated following vaccination

Many immune-relevant DEG were significantly downregulated following vaccination, with the most prominent immune genes being a CCL8–like gene (LOC126062469; FC: −60.8), the chemokines CXCL11, CXCL10, CCL24, CXCL6, CXCL1 (FC: −48.5, −23.7, −20.3, −15.4 and −9.4, respectively), the interferon-stimulated gene ISG15 (FC: −11.4), the pro-inflammatory cytokine IL1B (FC: −8.7) and the interferon-regulatory transcription factor IRF7 (FC: −7.6) (Supplementary Data 1a).

A high number of these DEG were associated with the innate immune system and the downregulation of innate immunity was corroborated by ORA of the DEG using the Human Gene Atlas. Here, particularly the CD14+ Monocyte gene set and a gene set relating to B cells (721 B lymphoblasts) were significantly downregulated (Fig. 3); and so were the CD56+ Natural killer (NK) cell gene set and a plasmacytoid dendritic cell-related gene set (BDCA4+ Dendritic cells), albeit to a lesser extent.

Gene expression in response to rEE2 and rMCP stimulation in vitro before and after vaccination

To further differentiate vaccine-induced responses from the latent EEHV immunity background of our study animals, contrast B (the pre-vaccination differential whole blood gene expression following rEE2 and rMCP antigen stimulation relative to time-matched PBS controls) was assessed first. Subsequently, these transcriptome changes could be compared to those in contrast C, which was generated to reveal the gene expression changes following antigen stimulation post-vaccination (Supplementary Fig. 1). In contrast B, 2957 DEG (padj ≤0.01) were present; a number that was substantially enhanced in contrast C to 6686 DEG (padj ≤0.01) (Supplementary Data 1b and 1c). There was an overlap of 2497 DEG between the pre-vaccination and post-vaccination dataset, of which 1341 genes were upregulated and 1155 genes downregulated. The number of upregulated genes in contrast B and C was 1489 and 3440, respectively.

The transcript showing the highest upregulation after antigen stimulation pre-vaccination was CXCL13 (FC: 3554), encoding a chemokine which preferentially attracts B lymphocytes and has been proposed as a plasma biomarker of germinal centre activity⁴³. A LIF-like gene (LOC126065436), IFNG and a IL36G-like gene (LOC126060847) ranked 5th to 7th among the top upregulated genes pre-vaccination (FC: 577, 524 and 512), respectively. Similarly, these genes were also among the top upregulated genes post-vaccination, where the IL36G-like gene was the top induced gene (FC: 6573) while IL1A was the 2nd most upregulated DEG (FC: 4511, rank 30 in contrast B) and CXCL13 was the 5th top upregulated gene (FC: 2517). IFNG represented the 4th top upregulated gene among the DEG in the post-vaccination dataset and was also induced with a higher FC (2665) than in contrast B. This demonstrated that IFNG was a good choice to determine antigen-specific reactions by RT-qPCR but also that the interpretation of single genes is particularly difficult in an immune memory situation.

Also among the top 25 induced genes following antigen-stimulation in both pre-and post-vaccination blood were IDO2 (FC B: 375; FC C: 1168) and IL1B (FC B:110; FC C: 980). Additional shared DEG present in the top 25 upregulated transcripts post-vaccination but ranking below the top 87 induced genes pre-vaccination included CCL22 (FC B: 36; FC C: 1221), CCL24 (FC B: 7; FC C: 431) and a GROG-like gene (CXCL3-like; LOC126077077; FC B: 24; FC C: 427).

In addition to these top ranked immune genes, a substantial number of other immune-relevant genes was detected among the remaining upregulated DEG in both pre- and post-vaccination transcriptomes. There was a considerable overlap between the chemo- and cytokine (receptor) signature induced during recall responses before and after vaccination, involving chemokine (receptors) CCL5, CCL20, CCL22, CCL24, CX3CL1, CXCL1, CXCL6, CXCL8, CXCL13, CCR7, CXCR2, CXCR4 and cytokine (receptors) CSF2, CSF3, IFNG, IL1A, IL1B, IL6, IL10, IL12A, IL12B, IL18, IL23A, IL27, IFNGR2, IFNLR1, IL1R1, IL2RA, IL4R, IL12RB2, IL21R, TNFSF13B, TNFSF14, TNFRSF4, TNFRSF9, TNFRSF10B, TNFRSF18, TNFRSF25, XCL1 and two LOC genes annotated as IL22 (LOC126075761) and IL22-like genes (LOC126075763).

Both datasets also shared interferon-stimulated genes ISG15 and ISG20 and several transcripts for cell surface proteins of immune genes including CD40, CD48, CD80, CD83, CD84, CD86, CD93, CD163, CD274 (PDL1), CD278 (ICOS), CD317 (BST2; tetherin) and immune-associated transcription factors and modulators such as AHR, ARID5B, ATF3, BACH2, BATF, BATF3, BHLHE40, ELF1, ETV3, IRF4, IRF7, JUNB, KLF4, KLF5, KLF6, NFIL3, NFKB1, NFKBIA, NFKBIE, NFKBIZ, PCNA, RBPJL, REL, RELB, RORA, RUNX1, RUNX3, SMAD1, SMAD7, SOX4, STAT4, ZBTB10, ZBTB18, ZBTB21 and ZBTB46.

Of note, the FC of the immune relevant DEG shared pre- and post-vaccination were usually higher or at a similar level post-vaccination, with few exceptions such as CXCL13 (the top upregulated gene before vaccination; FC B: 3554; FC C: 2517), complement component C8G (FC B: 50; FC C: 20), CX3CL1 (FC B: 34; FC C: 8), which can contribute to the recruitment of effector T cells to peripheral tissues and lymphoid organs⁴⁴ and participates in the adhesion between monocytes and endothelial cells⁴⁵, and IL12A (FC B:16; FC C: 11), a subunit of IL12 that plays an important role in the activities of T lymphocytes and NK cells.

Only a few genes with known or proposed immune functions were uniquely expressed before but not after vaccination. Of these, the decoy receptor IL1R2 (FC: 42) and a LOC126068746 predicted to encode a TLR13-like protein (FC: 22) ranked in the top 150 upregulated DEG. Further amongst the DEG that were solely upregulated pre-vaccination were the antiviral cytokine IFNB1, pro-inflammatory IL16 (functions as a chemoattractant and modulator of T cell activation⁴⁶), TNFRSF8 (expressed by activated T and B cells and has been shown to limit the proliferative potential of CD8+ effector T cells⁴⁷), the danger receptor TLR7 and a LOC gene (LOC126079100) predicted to code for a TRIM21–like protein (whereby TRIM21 is a putative modulator of IRF3 stability and a positive regulator of strength and duration of primary antiviral responses⁴⁸).

In contrast, a large set of 2099 unique DEG, among them a plethora of immune-related transcripts was upregulated after vaccination. These included, within the top 25 induced genes, IDO1 (FC: 1083), CXCL9 (FC: 484) and a CXCL8-like transcript (LOC126062469, FC: 447). Further immune genes detected within the top 150 DEG were MMP10 (FC: 314), CXCL11 (FC: 276), IL21 (FC: 238), CCL17 (FC: 171), CCL1 (FC: 122), a GAL10-like gene (LOC126085583; FC: 116), CLEC12B (FC: 90), IL17A (FC: 85), IL2 (FC: 76) and ICAM5 (FC: 70). Other uniquely present transcripts from different immune gene families included cyto- and chemokine (receptors) CXCL9, CXCL10, CXCL16, IL2, IL4, IL7, IL13, CCR4, CCR6, IL1RL1, IL2RG, IL10RB, IL12RB1, IL15RA, IL17A, IL18R1, IL20RB, IL27RA, TNF, TNFSF4, TNFSF10, TNFRSF1B, TNFRSF14, the CD markers and ligands CD1D, CD1E, CD2AP, CD4, CD28, CD38, CD40LG, CD44, CD53, CD68, CD70, CD74, CD109, CD226, CD247 and immune-relevant transcription factors and modulators such as ATF4, ATF5, ATF6, BBX, CITED2, CREB5, E2F3, E2F7, ELF2, EOMES, ETV6, ETV7, FOXO1, FOXP4, GATA3, GATA6, GFI1, JUN, KLF12, MAF, MAX, NFKB2, RBPJ, RELA, STAT1, STAT2, STAT3, STAT5B, TBX21 (TBET), a ZAP70-like gene (LOC126060770) and ZBTB1.

This vast expansion of uniquely upregulated DEG and the increases of FC of shared DEG seen post-vaccination demonstrated that a biologically meaningful interpretation of single genes was almost impossible and biased. Hence, we conducted an over-representation pathway analysis on antigen-specific upregulated genes pre- and post-vaccination to better understand the processes involved in latency protection and those stimulated or induced by this vaccine.

Over-representation analysis (ORA) using the gene ontology biological process (GOBP) database

Using the Gene Ontology Biological Process (GOBP) database first, ORA of significantly upregulated DEG in response to antigen stimulation before and after vaccination revealed that both conditions were enriched for a large amount of Gene Ontology (GO) terms with padj ≤0.05 and shared many of these terms (Supplementary Data 2a, b). More so, a large number of these GO terms was redundant with reference to the same processes, cell types or genes. Accordingly, we carried out a redundancy reduction approach to reduce the number of terms ultimately referring to the same set of genes (Supplementary Data 3a, b). The resulting lists of GO terms were analysed for those occurring pre-vaccination only, both pre- and post-vaccination or post-vaccination only (Supplementary Data 3c, d). Focussing on pathways directly linked to the immune system (Supplementary Data 3c), it became evident that pathways linked to an innate (NK cells) or early, inflammatory immune reaction (IL17) seemed to peak pre-vaccination (Fig. 4). A considerable number of GO terms, including antiviral pathways and those linked to the adaptive immune system (with emphasis on IFNG and IL12) were significantly enriched in both pre- and post-vaccination contrasts but gene overlaps were higher after vaccination. Additional terms referring to Th1 cells and unique terms relating to IL15 were present post-vaccination. Furthermore, two sets of GO terms were relatively novel upon vaccination: one set referring to the activation of B cells, the other to the innate immune system, particularly concerning the regulation of inflammatory responses and the production of IL1B (Fig. 4).

Over-representation analysis was performed using DEG identified from samples of two elephants (n = 2). Representative immune-relevant GO terms following reduction of redundant terms (>85% gene overlap) in contrasts comparing unstimulated vs. antigen-stimulated samples pre- and post-vaccination are shown according to their occurrence in the pre-vaccination comparison only, in both pre- and post-vaccination comparisons and post-vaccination comparisons only. DEG with an adjusted p value (padj) ≤0.01 were analysed. Significantly over-represented GO terms were identified using a one-sided Fisher’s exact test (H₀: no association), corrected for multiple comparisons with the Benjamini-Hochberg method, and ranked according to the combined score (z-score of the observed and expected means of the DEG and pathways overlap). The cut-off for consideration of GO terms as significant was padj ≤0.05. IFN interferon, IL interleukin, NK natural killer, Th T helper, TNF tumour necrosis factor.

ORA using the Reactome database

With GOBP providing a large, and in parts not very precise output of pathways, we also used Reactome as another database with immunological terms to review the immune activation in stimulated blood before and after vaccination. Here again the number of pathways identified increased significantly after vaccination (Supplementary Data 4a, b). For further consideration we used a cut-off of padj ≤0.01 and a combined score ≥50 pre-vaccination and ≥100 post-vaccination to focus on top regulated immunologically relevant pathways only.

In both contrasts, pathways relating to IFNG, but also to IL4/IL13 and IL12 were over-represented, demonstrating an adaptive immune response. Interestingly, also IL10 and IL35 pathways could be seen in common pointing to limiting immune responses, thus demonstrating a complete immune response both limiting the latent EEHV and extended upon vaccination.

In the post-vaccination contrast, a number of additional cytokine-related pathways were identified, specifically, IL18, IL21, IL23 and IL27. Also, we identified an enrichment of innate immunity-linked, specifically TNFR- and IL1-related pathways in Reactome too (Table 1).