Background
Influenza virus infections cause significant morbidity and mortality, with 5 million severely ill and 250 –500.000 deaths annually, in particular among the elderly, the immunocompromised and people with chronic diseases. The estimated global attack rate of influenza virus is 5 – 10 % for adults and 20 – 30 % for children, which causes large health and economic burdens for the society [
1]. Circulating seasonal influenza strains belong to A and B viruses. Influenza A viruses are classified on the basis of the antigenic properties of their hemagglutinin (HA) and neuraminidase (NA) glycoproteins. To date, 18 HA subtypes and 9 NA subtypes have been identified [
2]. Influenza B viruses are classified in two lineages, B/Yamagata and B/Victoria [
3]. While the host range of influenza B viruses is limited to humans and seals [
4], influenza A viruses infect a broad range of hosts including humans, birds and pigs [
5]. There is a constant threat of influenza A viruses crossing the species barrier and causing serious disease burden in humans [
6], as was recently demonstrated by human cases of avian H7N9 in China [
7].
Current trivalent seasonal influenza vaccines (TIV) are designed to elicit protective immunity against two specific influenza A strains (H1N1 and H3N2) and one B strain. The vaccines are mainly based on HA and primarily induce antibodies directed to the receptor binding site located on the globular head of the HA molecule which prevent the interaction of the virus with host cells and thereby block viral entry. However, since the globular head of the HA is highly variable [
8,
9] seasonal vaccines require annual updating to be effective. Each year the World Health Organization provides recommendations for the composition of seasonal influenza vaccines, based on predictions of the strains that will become dominant in the upcoming season. These predictions are based on global monitoring of the circulating H1N1, H3N2 and B strains, but this procedure is an error-prone process and mismatches between circulating virus and vaccine strains occur frequently. A particular challenge is posed by the fact that two B-strains (one from each lineage) are co-circulating. In the last decade, the dominant strain was correctly predicted in only 50 % of the cases [
10,
11].
Despite influenza strain variability, the HA contains conserved epitopes that may be targeted by vaccination [
12‐
15]. For the development of a broadly protective influenza vaccine that protects against mismatched seasonal strains and potential pandemic strains it may be beneficial to redirect the immune response towards such conserved epitopes.
A possible approach to this may be the use of adjuvants. It has been shown previously that adjuvants, such as MF59, the AS03 adjuvant system and saponin-based adjuvants have an ability to enhance and broaden the immune response elicited by vaccination [
16‐
26].
Here we evaluated the ability of a seasonal trivalent virosomal vaccine (TVV) adjuvated with the saponin-based adjuvant Matrix-M™ [
27,
28] to elicit heterologous hemagglutination inhibiting (HAI) antibodies and protection in mice. HAI antibodies are very potent at preventing the entry of influenza virus into the cell, through blocking the interaction between the HA head and cellular sialic acid receptors, and in this way fully prevent infection [
29,
30]. Furthermore, the HAI titer is the only accepted correlate of protection for influenza vaccines [
31].
We demonstrate that Matrix-M™-adjuvated TVV improves homologous HAI titers and protection. In addition, Matrix-M™-adjuvated TVV elicits heterologous HAI titers against influenza B, but not against influenza A, and provides protection against heterologous influenza strains in mice.
Discussion
Influenza vaccines need to be annually updated to efficiently protect against circulating strains. Therefore, it remains a high priority to establish broad reactive immunity against influenza by vaccination. An attractive approach may be to attempt redirecting the immune response towards the conserved epitopes on the viral antigens by adjuvated seasonal influenza vaccines.
It has previously been demonstrated that adjuvants can enhance and broaden immune responses of seasonal and pandemic vaccines in animal models and in humans [
16‐
24]. For example, MF59 (squalene oil-in-water emulsion [
33] has been shown to enhance vaccine-elicited humoral immune responses in animals [
20,
34,
35] and in humans [
35‐
39] and it is currently licensed for use in combination with a seasonal influenza vaccine (Fluad®). Fluad and an MF59 adjuvated H5N1 pandemic vaccine candidate are known to induce H1 and H5 cross-reactive HAI responses in ferrets [
20] and in humans [
16,
17,
36,
39], respectively. In adults, cross-reactive HAI responses were also detected against drifted H3N2 influenza viruses [
36]. MF59-adjuvated TIV also elicited some cross-reactive HAI responses against heterologous influenza B viruses in adults [
37] but not in unprimed children [
39].
Another adjuvant licensed for human use in combination with an H1N1 pandemic vaccine (Pandemrix®) is the AS03 adjuvant system (α-tocopherol and squalene in an oil-in-water emulsion [
40]). It has been demonstrated, in pre-clinical and clinical studies, that the AS03-adjuvated H5N1 pandemic vaccines can elicit H5 cross-reactive HAI responses which correlated to protection [
18,
19,
41].
In our studies, we have used Matrix-M™, an adjuvant that consists of a mixture of two purified and well characterized saponin fractions (Matrix-A and Matrix-C) [
27]. Compared to the original formulation that contained non-fractionated Quillaja saponins together with viral antigens in a single particle [
25], the individual particle formulation improves both adjuvant activity and the safety profile in humans [
21]. Matrix-M™ in combination with TVV has been shown to enhance antibody and cellular immune responses against influenza in mice [
22] and humans (Clinical-Trials.gov NCT01444482). Matrix-M™-adjuvated virosomal H9N2 and H5N1 vaccines have been evaluated in pre-clinical and clinical studies [
21,
23,
42‐
45]. Mice immunized with an experimental H9N2 virosomal vaccine showed enhanced homologous HAI responses [
45]. In addition, cross-clade H5 HAI responses were observed in mice that received a Matrix-M™-adjuvated H5N1 candidate vaccine compared to non-adjuvated H5N1 candidate pandemic vaccine in mice [
23]. These findings were later confirmed in humans [
21,
43].
Most studies concerning adjuvated influenza vaccines demonstrate cross-reactive responses against strains that are closely related to the vaccine strain.
Recently, we have shown that mice immunized twice with TVV + MM were protected from death after H5N1 and H7N7 challenge. Ferrets were partially protected against H5N1, but not against H7N9 after two immunizations with TVV + MM [
26].
In our study the H1N1 and H3N2 challenge strains only show 87 % and 80 % homology (based on HA protein sequence) with the corresponding vaccine strains and yet a single immunization with TVV + MM could protect mice against these influenza A strains.
The HA among influenza B strains is more conserved compared to influenza A (>91 % based on HA protein sequence as determined by alignment of the circulating strains from 1970 to 2008, data not shown). Nonetheless, the current TIV formulations are inadequate to elicit protection against both influenza B lineages, particularly in naïve individuals [
46,
47]. In addition, due to regular co-circulation of the strains from both lineages of influenza B, it is particularly difficult to predict which influenza B virus will be the dominant one in the upcoming season. The frequent mismatch between the influenza B lineage selected for seasonal vaccine and the dominant circulating B lineage results in the poor efficacy of seasonal vaccines against B strains [
10]. One solution to improve vaccine efficacy against B strains is the inclusion of strains from both influenza B lineages in the vaccine, the so-called quadrivalent influenza vaccine (QIV). An alternative approach may be the enhancement and broadening of the vaccine-elicited immune response by an adjuvant such as Matrix-M™. Our results demonstrate that a Matrix-M™ adjuvated seasonal trivalent virosomal vaccine elicits HAI titers not only against vaccine homologous B/Brisbane/60/08 but also against the vaccine mismatched B/Florida/04/06 and B/Massachusetts/02/12 (93 % homology with the vaccine strain for both) In addition to the enhanced protection against the vaccine matched B strain, full protection against a lethal challenge with the B/Florida/04/06 strain was achieved.
In contrast to influenza B, Matrix-M™-adjuvated TVV did not elicit cross-reactive HAI responses against heterologous H1N1 or H3N2 influenza A strains. The lack of cross-reactive influenza A HAI response is in agreement with our previous study [
26] and with studies that use a monovalent H1N1 split influenza vaccine formulated as ISCOMs [
24,
48].
The presence of cross-reactive influenza B HAI and the absence of cross-reactive influenza A HAI can be explained by the higher homology between the HA head of the two B lineages, as compared to the homology for H1N1 and H3N2 strains (Fig.
3b and Additional file
2: Figure S1, Additional file
3: Figure S2, Additional file
4: Figure S3). Similarly, cross-reactive HAI responses to relatively more closely related H5N1 strains were shown to be induced in mice immunized with a Matrix-M™-adjuvated virosomal H5N1 vaccine (98.7 % and 96.5 % HA homology of the NIBRG-88 and IBCDCRG-6 strains compared to the NIBRG-14 vaccine strain, respectively) [
23].
Considering that Matrix-M™-adjuvated TVV elicited HAI titers against two heterologous influenza B strains and provided full protection against challenge with the heterologous B/Florida/04/06, it is conceivable that HAI antibodies played a key role in protection, but does not exclude the contribution of other immunological mechanisms. Interestingly, despite the absence of HAI titers against the two heterologous influenza A strains, adjuvation with Matrix-M™ significantly improved TVV-induced protection against the heterologous H1N1 and H3N2 strains. In contrast to protection against heterologous influenza B, protection against influenza A was accompanied by the severe weight loss and disease symptoms. It is conceivable that the mechanism of heterologous protection induced by Matrix-M™-adjuvated TVV differs between influenza A and influenza B. We suggest that cross-reactive HAI antibodies play a key role in the protection against B/Florida/04/06. In the absence of cross-reactive HAI titers, as is the case for the two tested influenza A strains, other protective mechanisms are likely to play a key role. These include neutralization mediated by HA stem-binding antibodies [
12‐
14], antibody dependent cellular cytotoxicity (ADCC) [
24,
48‐
50], or T-cell mediated killing [
24,
48,
51,
52]. The exact mechanism of protection against heterologous influenza A strains induced by Matrix-M™-adjuvated TVV remains to be identified.
Materials and methods
Statement of ethics
All mouse and ferret experiments were performed in accordance with Dutch legislation on animal experiments and approved by the DEC Consult (Independent ethical institutional review board).
Immunization
Six-to eight-week-old female BALB/c (H-2d) mice (specific pathogen-free) were purchased from Charles River (France). H1N1 A/California/07/09, H3N2 A/Victoria/210/09 and B/Brisbane/60/08 monovalent virosomes were prepared by Crucell (Berne, Switzerland) using conventional procedures [
53]. The monovalent virosomes were mixed to obtain a trivalent virosomal vaccine (TVV). Matrix-M™ (MM, 10μg/dose, Novavax AB, Uppsala, Sweden) was mixed with TVV (at various doses, as indicated) before immunization (TVV + MM). Mice were immunized intramuscularly (i.m.) 1 to 3-times with TVV or TVV + MM 3 weeks apart with 100μl vaccine (50 μl per hind leg). Control groups received 100 μl PBS. Three weeks after the final immunization blood was collected to assess serum HAI responses (data shown for 1x and 2x immunization only). In the challenge experiments mice received mouse adapted influenza virus four weeks after the final immunization (data shown for 1x immunization only).
Hemagglutination Inhibition (HAI) assay
HAI assays were performed as described before [
45]. Briefly, sera were pre-absorbed with 0.5 % turkey red blood cells (bioTRADING Benelux B.V., Mijdrecht, the Netherlands) in PBS for 2 h. After removal by centrifugation, sera were treated for 16 h with receptor-destroying enzyme (Sigma-Aldrich; St. Louis, MO, USA; diluted 1:25 in PBS) The receptor-destroying enzyme was heat inactivated for 30 min at 56°C. Two-fold serial dilutions of the sera (initial dilution 1:8) were incubated for 1 h at room temperature with influenza virus (H1N1 A/California/07/09 reassortant (NYMC X-181), H1N1 A/Brisbane/59/07, H3N2 A/Perth/16/09, H3N2 A/Hong Kong/1/68, B/Brisbane/60/08 reassortant (NYMC BX-35), B/Florida/04/06, and B/Massachusetts/02/12) (see HA sequence alignment in Additional file
4: Figure S3, Additional file
5: Figure S4, Additional file
6: Figure S5) standardized to 8 hemagglutination units. Turkey red blood cells were added and incubated for 60 min before hemagglutination inhibition was determined. Each serum sample was tested in duplicate. Titers were expressed as the reciprocal of the highest dilution where complete agglutination inhibition was observed.
Influenza challenge
Four weeks after the final immunization mice were anesthetized by intraperitoneal (i.p.) administration of 100 mg/kg ketamine (Nimatek® 100 mg/ml, Eurovet, Cuijk, the Netherlands) in combination with 20mg/kg xylazine (Sedamun® 20 mg/ml, Eurovet). Mice were challenged with 25xLD
50 of (i) H1N1 A/Netherlands/602/09 or mouse-adapted H1N1 A/Brisbane/59/07 or B/Florida/04/06 (performed at Janssen Research and Development, Leiden, the Netherlands) or (ii) mouse-adapted H3N2 A/Perth/16/09, H3N2 A/Hong Kong/1/68 or B/Malaysia/2506/04 (performed at TNO Triskelion, Zeist, the Netherlands) (HA sequence alignment see Additional file
4: Figure S3, Additional file
5: Figure S4, Additional file
6: Figure S5) via the intranasal route (a total of 50μl, 25μl per nostril). Notably, all challenge strains were wild-type or wild-type derived and did not contain any PR8 derived segments.
After challenge, mice were monitored for weight-loss, clinical score and survival for 21 days or until humane endpoint. Clinical scores for challenges performed at Janssen Research and Development (i) were defined as: 0 = no clinical signs, 1 = rough coat, 2 = rough coat, less reactive, passive during handling, 3 = rough coat, rolled up, labored breathing, passive during handling, 4 = rough coat, rolled up, labored breathing, unresponsive (=humane endpoint) or found dead (=score 4).
Clinical scores for challenges at TNO (ii) were defined as 0 = no clinical signs, 1 = rough coat, 2 = rough coat, labored respiration 3 = rough coat, labored respiration, hunched posture and/or blepharospasm, 4 = rough coat, labored respiration, hunched posture, blepharospasm, lethargic and/or thin/dehydrated. Humane endpoint was defined as clinical score 4 for more than 2 consecutive days (=score 5) and found dead = score 5.
Statistical analysis
Statistical differences between immunizations with TVV with or without Matrix-M™ relative to negative control group receiving PBS were evaluated for HAI titers. Additionally, the effect of two immunizations compared to one immunization and the effect of Matrix-M compared to TVV alone were determined (Additional file
1: Table S1). Data was log-transformed and comparisons between groups were made using the Wilcoxon rank-sum test with adjustment for multiple comparisons (2 fold Bonferroni and for comparisons with PBS a stepwise approach testing first 2x and then 1x vaccination).
In the challenge models the vaccine groups were compared to the vehicle control group for survival proportion, change in body weight and clinical scores. For survival proportion after challenge, a Fisher’s exact test was performed. For body weight loss and clinical score analysis, repeated measurements in the challenge phase were summarized as a single outcome per animal using an Area Under The Curve (AUC) approach where missing values for animals that died before day 21 were imputed with a last-observation-carried-forward method. Body weight data are expressed as the change relative to the day 0 measurement. The AUC was then defined as the summation of the area above and below the baseline. An ANOVA on AUC’s was done with group as explanatory factor. Clinical scores were summarized as AUC per mouse and groups were compared using a generalized linear model with a cumulative logit distribution to compare area under the curves for ordinal variable.
For the vaccine homologous challenges, survival proportion per dose was compared to the vehicle control group using a 2-sided Fisher’s exact test. Across doses TVV + MM was compared to TVV only in a logistic regression model with log(dose) as covariable, adjuvant as factor and their interaction.
Statistical analyses were performed using SAS version 9.2 (SAS Institute Inc. Cary, NC, USA) and SPSS version 20 (SPSS Inc., IL, USA). Statistical tests were conducted two-sided at an overall significance level of α = 0.05.
Competing interests
All authors are or were employees of Janssen Research and Development, Pharmaceutical companies of Johnson and Johnson or Novavax AB.
Authors’ contribution
FC designed the experiments, interpreted the data and wrote the manuscript. ES designed the animal studies and helped interpret the data and draft the manuscript. MB acquired and analyzed the serological data and performed the animal studies. MK and JT supported the design of the animal studies and performed statistical analysis of the data. LD designed and coordinated the animal studies. WK and KLB helped to draft the manuscript. JG and KR designed the experiments, interpreted the data and helped to draft the manuscript. All authors read and approved the manuscript.