Effects of pre-existing orthopoxvirus-specific immunity on the performance of Modified Vaccinia virus Ankara-based influenza vaccines

The replication-deficient orthopoxvirus modified vaccinia virus Ankara (MVA) is a promising vaccine vector against various pathogens and has an excellent safety record. However, pre-existing vector-specific immunity is frequently suggested to be a drawback of MVA-based vaccines. To address this issue, mice were vaccinated with MVA-based influenza vaccines in the presence or absence of orthopoxvirus-specific immunity. Importantly, protective efficacy of an MVA-based influenza vaccine against a homologous challenge was not impaired in the presence of orthopoxvirus-specific pre-existing immunity. Nonetheless, orthopoxvirus-specific pre-existing immunity reduced the induction of antigen-specific antibodies under specific conditions and completely prevented induction of antigen-specific T cell responses by rMVA-based vaccination. Notably, antibodies induced by vaccinia virus vaccination, both in mice and humans, were not capable of neutralizing MVA. Thus, when using rMVA-based vaccines it is important to consider the main correlate of protection induced by the vaccine, the vaccine dose and the orthopoxvirus immune status of vaccine recipients.


VACV and H1N1pdm09 virus dose-finding. Sub-lethal doses of VACV and pandemic influenza virus
(H1N1pdm09) were determined in dose-finding experiments in C57BL/6 mice. Inoculation of mice with 10 4 -107 plaque forming units (PFU) VACV-Elstree by tail scarification led to weight loss (Fig. S1A), concurrent with the appearance of blisters at the site of inoculation in all mice (Fig. S1B). Similar levels of VACV-specific antibody responses were detected in all groups two weeks after inoculation (Fig. S1c). In addition, VACV-and MVAspecific CD4 + and CD8 + T cell responses were detected with a trend of stronger T cell responses at increasing infectious doses (Fig. S1D,E). A dose of 10 7 PFU VACV was considered the optimal sub-lethal priming dose for subsequent experiments.
In contrast to VACV, intranasal (IN) inoculation of mice with incrementing doses of H1N1pdm09 virus resulted in severe weight loss (Fig. S2A). Mortality was observed in mice inoculated with 10 5 and 10 6 tissue-culture infectious dose −50 (TCID 50 ) of H1N1pdm09 virus (Fig. S2B). Optimal induction of hemagglutination inhibition (HI) antibody responses (Fig. S2C) and T cell responses (Fig. S2D) without mortality was observed after inoculation with 10 4 TCID 50 , which was therefore considered the optimal dose for subsequent sub-lethal priming infections.
Induction of pre-existing orthopoxvirus-specific and influenza virus-specific immunity. According to the indicated priming regimens (Table 1, week 0 and/or 4) orthopoxvirus-specific or H1N1pdm09 influenza virus-specific immunity was induced. Four weeks after the last priming inoculation (week 8), induction of orthopoxvirus-or influenza virus-specific immunity was assessed by measuring serum antibody responses by protein array (PA) and ELISA. Priming with wtMVA or H1N1pdm09 influenza virus induced homologous antibody responses measured by PA (Fig. 1A). Serum antibodies reactive with wtMVA could not be detected in VACV-primed mice in this assay. Therefore, induction of VACV-specific antibodies by VACV priming was confirmed by ELISA (Fig. 1B). Notably, orthopoxvirus-or influenza virus-specific antibody responses were not detected in unprimed mice (Fig. 1A,B). In addition to detection of serum antibodies, H1N1pdm09 virus infection was confirmed by monitoring body weight of mice two weeks post-priming (Table 1, subgroup d). H1N1pdm09-virus inoculated mice lost body weight up to 7 days post-inoculation (dpi) and had regained their original weight at 11 dpi (Fig. 1C). In summary, priming with wtMVA, VACV or H1N1pdm09 was successful and induced detectable pre-existing immunity against the respective viruses in C57BL/6 mice. Pre-existing orthopoxvirus-specific immunity had limited effect on induction of antigen-specific antibody responses by rMVA. To determine the effect of pre-existing immunity on rMVA vaccine immunogenicity, unprimed and primed mice were vaccinated with rMVA expressing influenza virus nucleoprotein (NP) or HA (Table 1). Serum antibody responses against wtMVA and various HA1 subunits (HA from H1N1pdm09, H3N2 isolate from 2003 and 2011, and a selection of H5Nx viruses) after a single rMVA vaccination were determined by PA. As expected, rMVA vaccination consistently boosted the MVA-specific antibody response in mice primed with wtMVA or VACV ( Fig. 2A, compare wtMVA response of wtMVA-/VACV-primed mice with unprimed/H1N1pdm09 primed mice in all groups). Furthermore, boosting of H1N1pdm09-specific antibodies was observed in mice primed with H1N1pdm09 virus and subsequently vaccinated with rMVA-H1 ( Fig. 2A, group 3 and 7).
Induction of serum antibody responses against the corresponding antigen (H1pdm09 for rMVA-H1 and H3 [2003] for rMVA-H3 vaccination), was not hampered by either orthopoxvirus-or influenza virus-specific pre-existing immunity ( Fig. 2A). Antibodies against the heterologous H3 (2011) were not detected after a single vaccination with rMVA-H3. In contrast, a single rMVA-H5 vaccination led to induction of antibody responses to the homologous H5 antigen, as well as heterologous H5 antigens. In wtMVA-primed mice, lower antibody titers against the HA1 of all tested H5 clades, including the homologous A/Vietnam/1194/04 (H5N1, clade 1), was observed compared to unprimed, VACV-or H1N1pdm09-virus primed mice. This effect was particularly detected when a lower dose of rMVA-H5 (10 7 PFU, Table 1, group 6 at week 12) was used for the initial vaccination and to a lesser extent with the use of a higher vaccine dose (10 8 PFU, Table 1, group 5 at week 16) ( Fig. 2A,B). These results were confirmed by HI assay, which is a good proxy for influenza virus neutralization. Corresponding to the PA data, MVA-specific pre-existing immunity negatively affected the HI antibody response to influenza virus A/Vietnam/1194/04 after a single rMVA-H5 vaccination, especially when a low dose was used (Fig. 2C).  A second vaccination with rMVA (Table 1, week 16) boosted serum antibody responses to wtMVA and influenza viruses of interest. Similar to antibody responses induced by a single vaccination, antibody responses against the corresponding antigen after two vaccinations with rMVA-H1 or rMVA-H3 were not affected by pre-existing immunity (Fig. 3A,B, group 3-4). The second vaccination with rMVA-H3 induced cross-reactive antibody responses against an antigenically distinct H3 (2011). Notably, this cross-reactive response was detected in all subgroups, but was lower in mice with MVA-specific pre-existing immunity (Fig. 3A,B). Furthermore, in contrast to the antibody response after a single rMVA-H5 vaccination, the response after two rMVA-H5 vaccinations (Table 1, group 6) or an rMVA-H1 vaccination followed by 10 8 PFU rMVA-H5 (Table 1, group 7) detected by either PA or HI was not affected by pre-existing immunity to the vector ( Fig. 3A and C). Interestingly, recurrent vaccination with rMVA expressing different antigens (rMVA-H1 and rMVA-H5) still lead to efficient induction of antibody responses against both HAs. In conclusion, an effect of pre-existing MVA-specific, but not VACV-or influenza virus-specific, immunity on induction of humoral responses by rMVA vaccination was observed under specific conditions. Induction of antigen-specific T cell responses by rMVA is prevented by pre-existing orthopoxvirusspecific immunity. To determine the effect of pre-existing immunity on rMVA-induced antigen-specific T cell responses, splenocytes were obtained from unprimed and primed mice one or two weeks after the second rMVA-NP or rMVA-H5 vaccination, respectively ( Table 1, group 1-2). Antigen-specific CD8 + T cell responses were determined by measuring the number of interferon (IFN)-γ producing splenocytes after stimulation with Antigens (x-axis) and priming groups (y-axis) are indicated. Scale shows 2-log transformed titers. (B) VACV-specific serum antibody responses were determined by ELISA using VACVinfected HeLa cell lysate. The background signal on mock-infected cell lysate was subtracted. Individual sera from VACV-primed mice were used where possible. Serum from unprimed, wtMVA-primed or H1N1pdm09primed mice was pooled due to limited serum availability. Mean per priming group is indicated. Statistically significant differences were determined using a one-way ANOVA with multiple comparisons. ****p < 0.0001. (C) Mean body weight per group (n = 6, group 7 n = 5) after IN inoculation with 10 4 TCID 50 H1N1pdm09. No statistically significant differences between the groups were detected with a repeated measures ANOVA model. synthetic peptide NP 366-374 , an immunodominant CD8 + T cell epitope. Furthermore, the H5-specific CD4 + T cell response was determined after stimulation of splenocytes with full-length HA protein from H5N1 influenza viruses A/Vietnam/1194/04 (clade 1) or A/Indonesia/5/05 (clade 2.1).
rMVA-NP vaccination efficiently induced antigen-specific CD8 + T cell responses in unprimed and H1N1pdm09-primed mice but failed to induce NP-specific CD8 + T cells in mice with orthopoxvirus-specific pre-existing immunity, induced by either wtMVA or VACV priming (Fig. 4A). Similar observations were made in animals vaccinated with rMVA-H5: unprimed and H1N1pdm09 primed animals developed HA-specific CD4 + T cell responses against both the homologous (A/Vietnam/1194/04) and heterologous HA (A/Indonesia/5/05), but pre-existing orthopoxvirus-specific immunity had a detrimental effect on the induction of H5-specific CD4 + T cell responses (Fig. 4B).
Pre-existing orthopoxvirus-specific immunity impaired the protective efficacy of rMVA-based vaccines against challenge with a heterologous but not homologous virus. Four weeks after the final vaccination, rMVA-H1, rMVA-H3 and rMVA-H5 vaccinated mice were challenged with a lethal dose of influenza virus H5N1 (A/Vietnam/1194/04) in order to determine the effect of pre-existing immunity on the protective capacity of rMVA-based influenza vaccines (Table 1, group 3-7). As expected, rMVA-H1 and rMVA-H3 vaccination did not fully protect against an H5N1 influenza virus challenge, which was reflected by loss of body Mock-infected BHK-21 cell lysates were used as negative control for the wtMVA-infected cell lysates. Each horizontal line represents an individual animal grouped according to the vaccination group (rMVA-H1, rMVA-H3 or rMVA-H5) and priming subgroup (unprimed, wtMVA-primed, VACV-primed or H1N1pdm09-primed) indicated on the y-axis. Each vertical bar represents a different antigen (indicated on the x-axis). Scale represents 2-log titers as determined by PA, darker colors indicate higher antibody titers. (B) PA titers against H5 from A/Vietnam/1194/04 (clade 1) four weeks after rMVA-H5 vaccination with a high dose (10 8 PFU, group 5, week 16) or a low dose (10 7 PFU, group 6, week 12). Mean per priming group is indicated. Statistically significant differences were determined using a one-way ANOVA with multiple comparisons. *p = 0.0257, **p < 0.0017. (C) HI titers against influenza virus A/Vietnam/1194/04 were determined for each individual animal four weeks after rMVA-H5 vaccination with a high dose (10 8 PFU, group 5, week 16) or a low dose (10 7 PFU, group 6, week 12). Mean is indicated. Statistically significant differences were determined using a Kruskal-Wallis test. *p = 0.0401.
SCIENTIfIC RepoRts | (2018) 8:6474 | DOI:10.1038/s41598-018-24820-2 weight, lower survival rates and high viral loads in the lungs (Fig. 5). However, a limited level of cross-protection was observed after rMVA-H1 or rMVA-H3 vaccination in unprimed and H1N1pdm09-primed animals, which was not observed in mice primed with VACV or wtMVA (Fig. 5). Notably, pre-existing orthopoxvirus-specific or influenza virus-specific immunity did not interfere with the protective capacity of rMVA vaccines expressing the homologous HA gene of the H5N1 challenge virus since all mice that received at least one rMVA-H5 vaccination were fully protected from lethal H5N1 challenge (Figs 5 and S3).

Pre-existing MVA-specific, but not VACV-specific, antibodies have MVA-neutralizing capacities.
Whether antibodies induced by single or multiple MVA or VACV exposures have the capacity to recognize and neutralize MVA was investigated in vitro by ELISA and a virus neutralization assay. Mice that received a single exposure to either MVA or VACV did not induce any detectable MVA-specific serum antibody responses. In contrast, in mice that were exposed to MVA at least twice or once to VACV followed by at least one rMVA exposure, MVA-specific antibodies were detected. Notably, the MVA-specific antibody response was not boosted in response to additional exposures after the third MVA exposure (Fig. 6A). Similar observations were made in a VACV-specific ELISA, where at least two vaccinations with MVA or once with VACV followed by MVA led to detectable VACV-reactive antibodies (Fig. 6B). A booster effect of rMVA-based vaccinations after VACV priming was observed in both MVA-and VACV-specific antibody responses (Fig. 6A,B). Only sera obtained from mice exposed to MVA, but not from mice exposed exclusively to VACV, were capable of neutralizing MVA in vitro (Fig. 6C). Sera from humans shortly following MVA vaccination, but not 40 years after smallpox vaccination, are capable of neutralizing MVA. Next, the presence of orthopoxvirus-specific antibodies and their neutralizing capacity was assessed using serum obtained from humans vaccinated with either VACV ( ± 40 years post-vaccination) or rMVA (4 weeks after the third vaccination 20 ). MVA-specific antibody responses were detected four weeks after vaccination in individuals that received multiple vaccinations with 10 8 PFU rMVA-H5 20 (Fig. 7A). Furthermore, sera from VACV-vaccinated individuals (born between 1970-1971) and unvaccinated controls (born between 1976-1978) were probed for the presence of VACV-specific antibodies 30 . Almost four decades after vaccination, VACV-specific antibodies were still detected in VACV-vaccinated individuals, but not in the controls that were born 2-4 years after the smallpox vaccination campaign was terminated (Fig. 7B). Notably, sera obtained from rMVA vaccinated donors neutralized MVA efficiently in vitro, whereas sera obtained from VACV-vaccinated individuals were not capable of neutralizing an MVA infection (Fig. 7C). Although this is of course potentially due to waning of VACV-specific antibodies and does not necessarily reflect absence of cross-reactivity, this reflects the physiological situation at this time.

Discussion
Although it has been frequently suggested that pre-existing immunity to orthopoxviruses can interfere with the immunogenicity and efficacy of rMVA-based vaccines, the issue has not been addressed sufficiently in a well-controlled fashion. In this study, we investigated the performance of rMVA-based influenza vaccines in the presence or absence of pre-existing immunity to orthopoxviruses or influenza virus in mice and evaluated orthopoxvirus-specific immune responses after MVA or VACV vaccination in humans.
Induction of orthopoxvirus-specific antibodies upon priming mice with wtMVA or VACV was confirmed by PA and ELISA using wtMVA-or VACV-infected cell-lysates as antigens. Interestingly, VACV-specific antibodies did not cross-react with MVA in our PA assay, whereas MVA-specific antibodies clearly cross-reacted with VACV, as demonstrated by ELISA. These results are in accordance with a previous study 31 and similar results were obtained with human sera, where VACV-specific antibody responses cross-reacted with MVA only to a limited extent. In our study, the differential response can be explained by the fact that wtMVA-primed mice received a booster immunization inducing a MVA-specific recall response opposed to induction of a primary responses in the VACV-primed mice. Indeed, we observed that a single vaccination with rMVA expressing H1, H3 or H5 in unprimed mice did not induce detectable levels of MVA-specific antibodies, whereas rMVA vaccination of VACV-primed mice led to a boosting of MVA-specific antibodies. This corresponds to data obtained in humans, which show that MVA-specific antibodies were boosted upon MVA vaccination of either smallpox or rMVA vaccinated individuals 20,31 .
Pre-existing orthopoxvirus-specific immunity in mice affected humoral immune responses induced by rMVA influenza vaccines to a limited extent. Effects were exclusively observed in wtMVA-primed mice and only under 'suboptimal' conditions. When a lower rMVA-H5 vaccine dose was used, wtMVA-primed mice had reduced antibody titers to all tested H5 clades compared to the other priming groups. The reduced H5-specific antibody response observed in mice with MVA-specific pre-existing immunity was overcome by a second immunization with 10 8 PFU of rMVA-H5, as was previously reported 27 . Furthermore, in wtMVA-primed mice the response to an antigenically distinct H3N2 virus was significantly lower after two rMVA-H3 vaccinations compared to unprimed animals. However, pre-existing immunity had no effect on the magnitude of the antibody response to a corresponding H3 antigen induced by one or two rMVA-H3 vaccinations. Importantly, negative effects on the induction of antigen-specific antibody responses induced by rMVA were exclusively observed after wtMVA priming. VACV-specific pre-existing immunity never had any effect on the antigen-specific antibody response. In addition to antibody responses, the induction of antigen-specific T cell responses after rMVA vaccination in the absence or presence of pre-existing immunity was assessed. Virus-specific T cells contribute significantly to protective immunity against virus infections and can reduce duration and severity of disease [32][33][34] . Induction of T cells to influenza virus by vaccines is particularly attractive, since these mainly recognize epitopes in conserved internal proteins and can therefore afford cross-protection against various influenza viruses of different subtypes (heterosubtypic immunity) [35][36][37][38] . Our results indicated that induction of influenza virus-specific T cell responses was severely hampered by presence of MVA-or VACV-specific pre-existing immunity in mice. These findings are in concordance with previous studies that examined the immunogenicity of rMVA expressing human immunodeficiency virus (HIV) antigens in both mice 39 and macaques 28,40 in the presence of pre-existing vector-specific immunity. In contrast, a recent clinical trial reported efficient induction of cytomegalovirus (CMV)-specific T cell responses with rMVA in VACV-vaccinated individuals. The authors claimed that pre-existing VACV-specific immunity did not affect immunogenicity of rMVA, however, only a limited number of study subjects with pre-existing immunity was studied and their orthopoxvirus immune status was solely defined by date of birth and not confirmed with immunological assays 41 . Furthermore, rMVA expressing NP and matrix 1 (M1) genes from influenza virus has been shown to induce T cell responses in humans 42,43 , even in the elderly that potentially have orthopoxvirus-specific immunity 44 . Although this suggests that T cell responses can be induced in humans with VACV-specific immunity, the immune status of study subjects was not verified in these studies and appropriate control groups were lacking.
Pre-existing orthopoxvirus-or influenza virus-specific immunity did not affect survival of rMVA-H5 vaccinated mice after a lethal H5N1 influenza virus challenge. Protection was most likely mediated by antigen-specific neutralizing antibodies, which have been shown to be the main correlate of protection induced by this rMVA-H5 vaccine 20,45-48 , and were unaffected by the presence of pre-existing immunity at week 16. In contrast, mice that were vaccinated with either  rMVA-H1 or rMVA-H3 were partially protected from lethal H5N1 influenza virus challenge. A limited level of protection against H5N1 influenza virus infection was observed in unprimed or H1N1pdm09-primed mice, most likely mediated by cross-reactive antibody or T cell responses against influenza virus induced by priming and/or vaccination. Notably, mice with pre-existing orthopoxvirus-specific immunity had higher viral loads in the lungs and more severe weight loss compared to unprimed or H1N1pdm09 primed mice. In accordance with the described immunogenicity results, we hypothesize that orthopoxvirus-specific pre-existing immunity prevented the induction of antibody and/or T cell responses by rMVA-H1 or rMVA-H3 vaccination that are cross-reactive with H5.
It has been shown previously that VACV-or MVA-based vaccination efficiently induces both orthopoxvirusspecific antibodies and T cell responses 19,20,29,[49][50][51][52] (reviewed in 53 ). Hypothetically, vector-specific antibodies induced by previous immunizations could capture and neutralize rMVA virus particles upon (re-)vaccination, but non-neutralizing antibodies or orthopoxvirus-specific T cells could also play a role in interference. Interestingly, VACV-induced pre-existing immunity only interfered with induction of antigen-specific T cell responses but not antibody responses whereas MVA-induced pre-existing immunity could interfere with both. Since we have demonstrated that VACV-specific antibodies cannot cross-neutralize MVA in vitro, we hypothesize that interference of MVA-induced pre-existing immunity with induction of antigen-specific antibody responses is mediated by vector-specific antibodies. Similar results were obtained in humans, where VACV-induced antibodies could not neutralize MVA in vitro and only MVA-induced antibodies had MVA-neutralizing capacity. Follow-up adoptive transfers studies should be performed to identify the exact mechanism of interference of orthopoxvirus-specific immune responses with performance of MVA-based vaccinations.
It is important to note that our study in mice reflects a "worst-case scenario", since a time interval of only four weeks between induction of pre-existing immunity and initial vaccination with rMVA was maintained, not allowing for waning of orthopoxvirus-specific immunity. This does not accurately reflect the human situation, where smallpox vaccination was discontinued in the mid 1970s 18 . Even though VACV-specific antibody responses were still detected in the serum of vaccinated individuals by ELISA, these antibodies did not have MVA-neutralizing capacity in vitro. However, the timing used in this study does reflect the 'standard' interval used in rMVA vaccination regimens with multiple vaccinations, in which four-week intervals are frequently observed 11,20,41 . Our results show that repeated rMVA vaccination of humans does induce MVA-specific antibodies, which have neutralizing capacities in vitro and therefore may interfere with the immunogenicity of subsequent vaccinations.
In conclusion, the present study aids our understanding regarding immunogenicity of MVA-based vaccines in the presence of orthopoxvirus-specific immunity. Importantly, in mice rMVA is still immunogenic in the presence of orthopoxvirus-specific immunity, however, certain scenarios exist where pre-existing immunity can interfere with vaccine efficacy. This study represents an initial foundation to evaluate the effect of pre-existing immunity in a well-controlled fashion, future studies are warranted to elucidate the mechanisms underlying this interference. When using MVA-based vaccines, it is essential to consider the orthopoxvirus immune status of vaccine recipients, the interval between vaccinations in case of repeated rMVA-based vaccination, the vaccine dose used and the main correlate of protection induced by rMVA-based vaccines to ensure induction of an optimal immune response.

Materials and Methods
Ethics statement. Animal experiments were conducted in strict compliance with European guidelines (EU directive on animal testing 2010/63/EU). The animal protocol was approved by an independent animal experimentation ethical review committee (Erasmus MC permit number EUR3277-02). Animal welfare was observed on a daily basis, and all invasive animal handling was performed under anaesthesia using 4% isoflurane in oxygen to minimize animal suffering. Human sera pre-and post-rMVA vaccination (three vaccinations at week 0 and 56 with 10 8 PFU rMVA-H5, n = 6) were obtained during a randomized, double-blind phase 1/2a study at the Erasmus MC, Rotterdam, the Netherlands. The study involved adult volunteers (male/female, between ages 18-28) who provided informed consent. The study design was reviewed and approved by the Central Committee on Research involving Human Subjects in the Netherlands 20 . Furthermore, serum samples from VACV vaccinated (n = 6) and unvaccinated (n = 6) healthy individuals (male/female) were collected during a cross-sectional population-based study performed in the Netherlands from February 2006 until June 2007 (PIENTER2 study) 30 . Smallpox vaccination campaigns lasted until September 1974 in the Netherlands. To limit the inevitable age bias, sera from individuals born between 1970-1971 and 1976-1978 were selected for the VACV vaccinated and unvaccinated group, respectively. The work described here has been carried out in accordance with the code of ethics of the world medical association (declaration of Helsinki).
Priming, rMVA vaccination and challenge. Mice (6-8 weeks old) were divided into seven groups (n = 24) with four subgroups (n = 6) each (Table 1). Animals were either unprimed (subgroup a) or were primed with 10 8 PFU wtMVA in 100 μl PBS intramuscularly (IM, two immunizations at week 0 and 4, subgroup b), 10 7 PFU VACV-Elstree (week 4, subgroup c) or 10 4 TCID 50 H1N1pdm09 (week 4, subgroup d). VACV-Elstree and H1N1pdm09 were administered as described above at the optimal priming dose determined in the dose-finding experiments mentioned above. After priming, mice received one or two IM vaccinations at week 8 and/or 12 with 10 8 PFU of rMVA-NP, rMVA-H1, rMVA-H3 and/or rMVA-H5 in 100 μl PBS. Of note, 10 7 PFU rMVA-H5 was administered at week 8 to the mice of group 6 ( Table 1) to establish if pre-existing immunity affects low dose rMVA-HA vaccination and if a boost with a high dose could overcome potential negative effects. After vaccination, mice vaccinated with rMVA-NP or rMVA-H5 were not challenged and were euthanized one (week 13) or two weeks (week 14) after the second vaccination, respectively (Table 1, group 1-2). The remainder of the animals ( Table 1, group 3-7) were challenged IN with 10 3 TCID 50 A/ Vietnam/1194/04 (H5N1) influenza virus four weeks after the second vaccination (week 16) and monitored twice daily. Mice were euthanized when pre-defined humane endpoint criteria (>25% body weight loss) were reached or at seven days post-challenge (week 17). Blood, spleen and/or lung samples were harvested during necropsy. A single mouse in the wtMVA-primed subgroup of the 2 × rMVA-H5 challenge group (group 6, subgroup b) had to be euthanized due to a wound unrelated to the experiment at week 4. One mouse in the H1N1pmd09-prime group of the 1 × rMVA-H1 and 1 × rMVA-H5 group (group 7, subgroup d) had to be euthanized 10 days after priming because humane endpoint criteria were met. These mice were excluded from further analysis. Animal experiments with groups 3 & 7 and 5 & 6 were performed in parallel but at different stages, groups 1 & 2 and group 4 were performed separately. Analysis of the samples was performed collectively for all groups.
Virus isolation from lungs. Directly after necropsy, all lungs were snap frozen and stored at −80 °C for processing at a later time point. To perform virus isolations, lungs were thawed, lung weight was recorded and lungs were homogenized with a Polytron homogenizer (Kinematica AG) in MDCK infection medium (without FBS). Quadruplicate ten-fold serial dilutions of these samples in MDCK infection medium supplemented with 0.002% TPCK-Trypsin (Lonza) were used to determine the virus titers on MDCK cells as described previously 57 . Stimulation and intracellular cytokine staining of splenocytes. During necropsy spleens were collected in Iscove's Dulbecco's Medium (IMDM, Lonza) supplemented with 5% FBS and P/S/G for direct preparation of single cell suspensions using 100 µm strainers (Falcon). Erythrocytes were removed from single cell suspensions by treatment with red blood cell lysis buffer (Roche diagnostics). For intracellular cytokine staining, splenocytes were stimulated with 5 µM synthetic peptide (epitope NP 366-374 : ASNENVEIM [ Fig. S2] or ASNEMMETM [ Fig. 4]) or 1 µg/250,000 cells recombinant HA protein from H5N1 influenza virus A/Vietnam/1203/04 or A/Indonesia/5/05 (Protein Sciences) in IMDM supplemented with GolgiStop and incubated for 6 h at 37 °C. Mock-treated splenocytes and splenocytes stimulated with 50 ng/ml PMA (Sigma-Aldrich) and 0.5 µg/ml ionomycin (Sigma-Aldrich) served as appropriate negative and positive controls. After stimulation, splenocytes were incubated with fluorochrome-labeled antibodies to CD3e APC-Cy7 (BD Pharmingen), CD8b FITC (BD Pharmingen), CD4 PerCP (BD Pharmingen) and viable cells were identified with Aqua LIVE/DEAD (Invitrogen). Subsequently, cells were fixed and permeabilized using BD Cytofix/Cytoperm TM Plus (BD Biosciences), and incubated with anti-IFN-γ PacificBlue (Biolegend). Samples were acquired on a FACS Canto II and data was analyzed as described previously 55,59 using FACS Diva software (BD Biosciences).
Protein Array (PA) assay. Mouse sera collected at week 8 and 16 (Table 1) were used to determine the presence of antibodies to selected antigens by PA as described previously 60 (Scienion). Sera were incubated on the slides in Blotto Blocking Buffer (Thermo Fisher Scientific Inc.) supplemented with 0.1% Surfactant-Amps (Thermo Fisher Scientific Inc.). Subsequently, goat-anti-human IgG labelled with AlexaFluor647 (Jackson ImmunoResearch Laboratories Inc.) was used as conjugate and fluorescent signals were measured using a Powerscanner (Tecan Group Ltd). The titer of each serum sample was defined as the interpolated serum concentration generating the 50% point using a four-parameter logistic nonlinear regression model using R (R Statistical Computing, version 3.1.0. Measured titers were corrected for the positive control included on each slide.

Detection of MVA-or VACV-specific antibodies by ELISA. For detection of VACV-specific antibodies,
HeLa cells were mock-treated or infected with VACV-Elstree at MOI 1 and harvested in 1% Triton-X100 in PBS supplemented with mini cOmplete EDTA free protease inhibitor tablet. Similar procedures were used to obtain BHK-21 cell lysates mock-treated or infected with wtMVA. Sera used for detection of MVA-or VACV-specific antibodies were pre-cleared O/N at 4 °C in a 96-well plate with confluent BHK-21 or HeLa cells, respectively. For ELISA, 96-well plates (Corning Costar) were coated overnight at 4 °C with 10-25 µl cell lysate (mock-treated or infected) per well in 0.05 M Carbonate/Bicarbonate pH 9.6. Plates coated with cell lysate were washed and all plates were blocked for 1 h at room temperature (RT) with blocking buffer consisting of PBS supplemented with 0.05% Tween-20 (PBST, Merck-solutions) and 2% milk powder (w/v, Campina). Subsequently, a 3-log dilution series of serum in blocking buffer was prepared, starting dilution 1:10 or 1:30, and 50 μl was transferred to wells of the antigen-coated plates and incubated for 1-1.5 h at RT. Blocking buffer and VACV-positive serum served as appropriate negative or positive control, respectively. Plates were washed with PBST and incubated for 1 h at RT with HRP-conjugated goat-anti-mouse IgG (DAKO) or goat-anti-human IgG (Southern Biotech). Plates were washed with PBST and incubated for 10 min with 50 µl TMB peroxidase substrate (KPL) after which the reaction was stopped with 0.5 M H 2 SO 4 (Merck). Absorbance was measured at 450 nm using a Tecan infinite F200. The OD 450 values at a single dilution in the linear area of the curve were determined and analysed. Due to limited amounts of serum, sera were pooled per subgroup (Table 1, n = 3-6). Sera of VACV-primed animals were tested individually in VACV ELISA as much as possible, although in some groups a few samples had to be pooled (indicated in figure legends). The OD 450 value obtained with mock-infected BHK-21 or HeLa cell lysate was subtracted from the OD 450 value obtained with the respective infected cell lysate to determine a net OD 450 response.
Hemagglutination inhibition (HI) assay. Sera were treated with a receptor-destroying enzyme (cholera filtrate) overnight at 37 °C, followed by heat-inactivation for 1 h at 56 °C. HI assay was performed in a 2-fold serial dilution in duplicate following a standard protocol using 1% turkey erythrocytes and four HA units of an H5N1 reverse genetics influenza virus with HA (without multibasic cleavage site) and NA gene segments of A/ Vietnam/1194/04 and the remaining gene segments of influenza virus A/Puerto Rico/8/34 (6 + 2) 62 .
Plaque reduction assay. Mouse sera were pooled per subgroup (Table 1, n = 2-5) due to limited serum availability. Sera were heat-inactivated for 30 min at 56 °C. A 2-log dilution series -starting dilution 1:10 -was prepared in CEF culture medium and incubated for 2 h with 200 PFU/well wtMVA (mouse sera) or rMVA-GFP (human sera) in a 1:1 ratio at 37 °C. Human serum with antibodies against MVA was used as a positive control. Subsequently, the serum-virus mixture was incubated for 2 h at 37 °C on a confluent monolayer of CEF cells in 96-wells culture plates. Cells were washed with PBS and incubated for 44-48 h 37 °C. CEF used for serology of mouse samples were fixed with acetone/methanol (Sigma-Aldrich). Plates were blocked with 3% FBS in PBS for 1 h. Subsequently, plaques were stained with rabbit anti-VACV (Lister strain, Acris) followed by goat-anti-rabbit HRP-conjugate (Jackson ImmunoResearch Laboratories Inc.). Samples were developed using True Blue (KPL). The percentage of area covered by stained plaques was measured using a CTL immunospot reader with CTL biospot software. CEF used for serology of human samples were fixed with 2% paraformaldehyde (PFA) for 10 min after which fluorescent plaques were detected using a Typhoon ™ FLA9500 (GE Healthcare). Plaques were counted using ImageQuant TL Colony v8.1 software (GE Healthcare). The MVA-neutralization titer was determined as the reciprocal of the highest dilution at which the area covered by plaques was below 50% of the average percentage of the area covered (mouse sera, Fig.  S4A) or counted spots (human sera, Fig. S4B) in n = 12 wells without any added serum.

Statistical analysis.
Longitudinal body weight data after H1N1pdm09 priming was analyzed using a repeated measures ANOVA model, with time as within factor. One way ANOVA with multiple comparisons was used to compare the normally distributed (according to the Shapiro-Wilk test) VACV-specific antibody responses after priming (week 8), H5-specific PA antibody responses after one vaccination and the 2003/2011 H3-specific PA antibody responses in mice. A Kruskal-Wallis test was used to compare the not normally distributed HI titers against A/ Vietnam/1194/04 after a single vaccination, NP-or H5-specific T cell responses and viral lung titers. HI titers below the detection limit (titer 40) were set to a titer of 20 (the highest possible titer below 40). Statistical differences in the MVA-specific PA response were determined relative to the '0 × MVA' control sample using a Kruskal-Wallis test. Furthermore, the ELISA and neutralization titers in human serum samples were compared using a Wilcoxon matched-pairs signed rank test (MVA sera) or a Mann-Whitney test (VACV sera). Neutralization titers in the plaque reduction assay below the detection limit (10) were set to a titer of 5 (the highest possible titer below 10).